Imaging Pearls ❯ July 2026
|
-- OR -- |
|
3D and Workflow
- There is a growing awareness that body CT scans contain rich cardiometabolic information that can be leveraged for additional patient benefits. However, the clinical implementation of opportunistic CT screening in routine practice has been hindered by valuable yet onerous manual measurements and subjective assessments. Explainable artificial intelligence (AI) algorithms are now poised to change this. The potential impact of opportunistic screening is further enhanced by the large volume of CT scans being obtained. In this “How I Do It” installment, the authors briefly outline some current approaches that can be obtained “on the fly,” while focusing more on emerging automated solutions. Detecting unsuspected or presymptomatic conditions, such as osteoporosis, cardiovascular disease, sarcopenia, and hepatic steatosis, could lead to preventive interventions, regardless of the original indication for imaging. Composite models that combine multiple cardiometabolic CT biomarkers can be applied to survival prediction and assessment of biologic aging, frailty, cancer cachexia, metabolic syndrome, and fracture risk, among other factors. For clinical reporting, a range of logistical, actuarial, and ethical issues must be carefully considered. However, if executed properly, we believe that opportunistic CT screening can add substantial value, be cost saving, and provide a new level of personalized precision medicine befitting the dawning AI information era.
CT-based Opportunistic Screening for Adding Clinical Value: How I Do It
Perry J. Pickhardt • Mathew H. Lee • Joshua D. Warner • Ronald M. Summers • John W. Garrett
Radiology. 2026 Apr;319(1):e252106. doi:10.1148/radiol.252106. - ■ Valuable incidental tissue and organ data in CT scans often go unused in clinical practice but can and should be responsibly repurposed for patient benefit through opportunistic screening.
■ Silent presymptomatic cardiometabolic conditions may be uncovered at CT imaging performed for other indications.
■ Diagnostic and predictive opportunistic results can be derived from understandable CT measures, without the need for opaque “black box” artificial intelligence (AI) models and radiomics.
■ Explainable AI algorithms can efficiently quantify these clinically relevant CT biomarkers, replacing more onerous manual measures.
■ Opportunistic screening reflects the shift from volume-based care in radiology toward a value-added focus. - Sarcopenia refers to the loss of muscle mass, strength, and function. Although part of normal aging, other factors can accelerate this process pathologically. Historically, the primary focus has been on low muscle mass (myopenia); however, we have repeatedly found that low muscle attenuation (myosteatosis), a measure of muscle quality, carries substantially more prognostic value. When assessing muscle attenuation, we include the intermuscular adipose tissue within the segmented region of interests.
CT-based Opportunistic Screening for Adding Clinical Value: How I Do It
Perry J. Pickhardt • Mathew H. Lee • Joshua D. Warner • Ronald M. Summers • John W. Garrett
Radiology. 2026 Apr;319(1):e252106. doi:10.1148/radiol.252106. - It is not hyperbole to suggest that CT already represents one of the greatest advances in medicine over the past half century. By leveraging the rich omnipresent cardiometabolic data incidental to the indication for imaging, opportunistic CT screening can add substantial value to personalized patient care, as well as to broader public health.
CT-based Opportunistic Screening for Adding Clinical Value: How I Do It
Perry J. Pickhardt • Mathew H. Lee • Joshua D. Warner • Ronald M. Summers • John W. Garrett
Radiology. 2026 Apr;319(1):e252106. doi:10.1148/radiol.252106.
- Pancreatic ductal adenocarcinoma (PDAC) presents as a cancer with an especially poor prognosis, largely due to the challenges surrounding its early diagnosis. Liquid biopsy has emerged as a promising, noninvasive method for screening across a variety of cancers. This approach is limited, however, by the extensive heterogeneity of biological samples, a challenge that teams are looking to address using artificial intelligence (AI) and machine learning (ML) strategies. By harnessing the ability of ML algorithms to extract the most salient features from complex datasets, researchers can identify biomarkers with high predictive value for PDAC. This review explores the current landscape of AI-powered liquid biopsy for early PDAC diagnosis, focusing on specific techniques and their respective degrees of success.
Machine learning and artificial intelligence in liquid biopsy based early detection of pancreatic cancer: a scoping review
Joy Ku, Meenakshi Singhal, Margaret Burnette and Samar A. Hegazy
BJC Reports; https://doi.org/10.1038/s44276-026-00232-y - miRNAs are small, non-coding RNA molecules that can be found in EVs, bound to other proteins, or circulating freely in biological fluids. They regulate gene expression post-transcriptionally, typically either by direct degradation of target mRNA or by reducing translation. Specific miRNAs can be dysregulated in cancer cells, and their stability within biological fluids means that they can serve as promising biomarkers for the detection of cancer-specific expression patterns, especially within PDAC.
Machine learning and artificial intelligence in liquid biopsy based early detection of pancreatic cancer: a scoping review
Joy Ku, Meenakshi Singhal, Margaret Burnette and Samar A. Hegazy
BJC Reports; https://doi.org/10.1038/s44276-026-00232-y - CONCLUSION
Non- or minimally invasive screening via liquid biopsy for PDAC is a promising and viable complementary tool to existing imagingbased screening measures in order to mitigate the procedural burden on patients. Given the tendency of PDAC to remain asymptomatic for extended periods of time, the future of its early detection requires more proactive measures, such as regular surveillance. AI/ML unlocks the potential for multianalyte screening panels that could provide more robust diagnostic value compared to CA19-9 alone. With continued improvements, a PDAC panel may eventually be routinely available for patients, which can improve patient outcomes. By reducing the barrier to accessing timely screenings, it may be possible to reverse PDAC’sreputation as one of the most lethal common cancers in the US.
Machine learning and artificial intelligence in liquid biopsy based early detection of pancreatic cancer: a scoping review
Joy Ku, Meenakshi Singhal, Margaret Burnette and Samar A. Hegazy
BJC Reports; https://doi.org/10.1038/s44276-026-00232-y
- With the rapid advancement of multi-detector computed tomography (MDCT) and image post-processing technologies, CT angiography (CTA) has become a cornerstone in the diagnosis of vascular diseases. Despite its widespread use, conventional volume rendering (VR) is limited in depth perception, visualization of fine tissue textures, and differentiation of complex, overlapping anatomical structures. Cinematic rendering (CR), a next-generation visualization technique, utilizes a global illumination model to achieve photorealistic 3D reconstructions.
Cinematic rendering in CT angiography: a pictorial review of clinical applications
Chong-ze Yang, Hua-song Cai, Lei Ding et al.
Insights into Imaging (2026) 17:139 - ● Cinematic rendering employs global illumination and high dynamic range to surpass traditional volume rendering, enhancing anatomical realism for effective vascular diagnosis and lesion detection.
● Cinematic rendering transforms complex data into intuitive three-dimensional visualizations, reducing cognitive load for clinicians and facilitating both medical education and patient communication.
● Integration with artificial intelligence and mixed reality holds the potential to advance cinematic rendering from morphological displays to intelligent, interactive tools for preoperative planning.
Cinematic rendering in CT angiography: a pictorial review of clinical applications
Chong-ze Yang, Hua-song Cai, Lei Ding et al.
Insights into Imaging (2026) 17:139 - Accurate preoperative anatomical assessment is pivotal for surgical success. CR technology, with its superior depth perception, serves as an important link between radiological imaging and surgical practice in preoperative tumor planning. Traditional 3D reconstructions often present with rigid textures, rendering the anatomical relationships between lesions and surrounding vessels as mere planar superimpositions that lack stereoscopic depth. Through meticulous post-processing, CR imaging not only facilitates visualization of tumor tissues with fine, anatomically realistic textures but also clearly delineates the spatial relationships between tumors and adjacent vasculature, including abutment, encasement, and invasion .
Cinematic rendering in CT angiography: a pictorial review of clinical applications
Chong-ze Yang, Hua-song Cai, Lei Ding et al.
Insights into Imaging (2026) 17:139 - CR provides high-fidelity, photorealistic threedimensional visualization in CTA, significantly enhancing the depiction of vascular morphology, pathologies of the vessel wall, and dynamic cardiac motion. Compared to conventional VR and MIP, CR improves depth perception tissue texture, and the spatial delineation of complex vascular structures. This supports accurate preoperative planning, postoperative assessment, and comprehensive evaluation of vascular disease. Moreover, CR datasets can also be integrated with artificial intelligence, mixed reality,and 3D printing to facilitate advanced image analysis, intraoperative guidance, and patient-specific modeling. These capabilities establish CR as a versatile platform precision medicine. However, further clinical validation and integration into routine workflows are necessary to fully realize its potential in improving diagnostic accuracy,surgical outcomes, and multidisciplinary decision-making.
Cinematic rendering in CT angiography: a pictorial review of clinical applications
Chong-ze Yang, Hua-song Cai, Lei Ding et al.
Insights into Imaging (2026) 17:139 - Artificial intelligence is fundamentally transforming the paradigm of medical image analysis, yet its performance is highly dependent on the quality and information density of the training data. Traditional CT images and VR often lack or lose textural information, which limits the ability of AI models in fine recognition and prediction tasks. Conversely, CR introduces lighting models that closely mimic physical reality, preserving minute textural features and complex spatial information. This high-quality data input holds the potential to significantly improve the AI models’ performance in detecting and predicting minute lesions. It is crucial to ensure “anatomical authenticity” and avoid introducing non-biological artifacts, which remain essential prerequisites for applying CR data to AI research. With these considerations in mind, the synergy between CR and AI is expected to enhance the detection capabilities and predictive performance of AI models.
Cinematic rendering in CT angiography: a pictorial review of clinical applications
Chong-ze Yang, Hua-song Cai, Lei Ding et al.
Insights into Imaging (2026) 17:139
- Primary small bowel malignancies are rare, often presenting with nonspecific symptoms or as acute emergencies, which can delay diagnosis. Contrast-enhanced CT is the primary imaging modality in the emergency setting, but detection and characterization of small bowel tumors remain challenging. Cinematic rendering (CR) is a recently developed three-dimensional post-processing technique that produces photorealistic images from CT data, enhancing visualization of small bowel pathology. This pictorial review outlines the CT imaging features of major small bowel malignancies, including adenocarcinoma, carcinoid tumor, gastrointestinal stromal tumor, lymphoma, and sarcoma, and describes features that highlight the utility of CR in augmenting traditional imaging. CR offers improved visualization of mucosal abnormalities, tumor extent, vascular involvement, and textural differences, potentially increasing diagnostic confidence, supporting presurgical planning, and facilitating communication among clinicians and patients. By emphasizing the added value of CR, we aim to provide radiologists with practical guidance for identifying small bowel neoplasms and suggest that integrating advanced 3D visualization into routine CT evaluation can support timely diagnosis and management in acute care settings.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Small bowel malignant neoplasms are rare, accounting for only 0.6% of all cancers and 3% of gastrointestinal (GI) malignancies in the U.S., though incidence has risen 118% over the past four decades. Among these tumors, small bowel adenocarcinomas and carcinoid (i.e. neuroendocrine) tumors together comprise approximately 80% of cases, with each accounting for about 40%. The remaining 20–25% are made up of other malignancies, including gastrointestinal stromal tumors (GIST), lymphomas and sarcomas . Several risk factors have been identified for the development of these malignant neoplasms, such as inflammatory conditions like inflammatory bowel disease and celiac disease; hereditary syndromes including familial adenomatous polyposis, hereditary nonpolyposis colorectal cancer, and Peutz-Jeghers syndrome; and infections such as HIV. .
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Additionally, photorealistic 3D images intrinsic to CR can help increase the understanding of these pathologies for students, trainees, and patients. At our institution, cinematic rendering is performed using syngo.via software. Predefined transfer functions within the CR platform map attenuation values from the original CT dataset to specific color palettes and opacity levels, thereby controlling tissue transparency. By selecting and fine-tuning different transfer functions and adjusting clipping planes, varying degrees of translucency can be achieved to selectively emphasize specific tissues and anatomical structures, enabling the generation of high quality, clinically meaningful CR images. The authors point out that for advanced 3D postprocessing methods like CR to be used effectively in clinical practice, institutions must have high quality contrast-enhanced images, access to the specialized software, and radiologists with specific training in these technologies.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Small bowel adenocarcinoma most commonly arises in the distal duodenum or proximal jejunum through malignant transformation of glandular cells, often originating from precursor adenomas. Tumors under 2 cm are commonly overlooked, especially because bowel dilation or obstructive features and clinical symptoms are uncommon. On CT imaging, small bowel adenocarcinomas usually appear as enhancing lesions that cause irregular, circumferential or eccentric narrowing of the bowel lumen. This can lead to obstructive symptoms, and a study of 217 patients with small bowel adenocarcinoma found emergency diagnoses due to occlusion in 40% and bleeding in 24% of patients.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Carcinoid tumors of the small bowel often present with vague symptoms such as abdominal pain, but may initially be identified by carcinoid syndrome which is marked by secretory diarrhea, flushing, telangiectasia, bronchial constriction, and potential cardiac valve abnormalities secondary to bioactive substances such as serotonin secreted by carcinoid tumors associated with liver metastases. These tumors originate from chromaffin cells at the base of the crypts of Lieberkühn, most commonly in the distal ileum. They generally grow as submucosal nodules and typically demonstrate intense early hyperenhancement on imaging. Detecting these lesions can be challenging with conventional imaging techniques, given their often sub-centimeter size and submucosal location.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Desmoplastic reaction may result in angulation, tethering, and fixation of the affected small bowel loops. Mesenteric involvement and desmoplastic reaction can be clearly visualized with CR by adjusting windowing levels, allowing for detailed depiction of features such as the ‘spoke-wheel’ or stellate pattern, as well as assessment of adjacent vessels and any vascular compromise. With CR, physicians can intuitively visualize the extent and locations of multiple lesions in a single 3D view. Another finding may be carcinoid metastases, most commonly found in the liver, where their hypervascular nature leads to bright enhancement on CT and CR . Additional metastatic sites can include the lungs, bones, and peritoneum.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - “On CT imaging, GISTs are typically seen as large, lobulated, well-circumscribed, and predominantly exophytic soft tissue masses with heterogeneous enhancement patterns, although small GISTs usually appear as sharply margined, smooth-walled, homogeneous masses. While calcifications are rare, neovascularity, central necrosis, ulceration, hemorrhage, or cavitation may be observed within the masses. Consequently, CR can be particularly helpful in evaluating the internal architecture of GISTs and identifying key features, due to its capability to accentuate textural differences.”
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Primary gastrointestinal lymphoma is the most common extranodal lymphoma, most frequently affecting the ileum due to its rich lymphoid tissue, and is predominantly of B-cell origin, with only 8–10% of cases arising from T-cells. Patients can present in emergency settings due to GI perforation, occurring in about 9% of lymphoma cases, with 59% involving the small bowel, although obstruction remains atypical due to the absence of a desmoplastic reaction. Along with chronic inflammatory conditions, infections are risk factors and include Helicobacter pylori, HIV, Campylobacter jejuni and Epstein-Barr virus among others. Radiologically, key indicators of primary small bowel lymphoma include enhancing wall thickening, with adjacent multiple enlarged mesenteric lymph nodes without necrosis.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Small bowel lymphoma can be ill-defined and has various forms and can present as pseudoaneurysmal wall thickening, polypoid intraluminal masses causing intussusception, endoexoenteric cavitary masses, exophytic bulky masses invading the mesentery, or rarely, stenosing fibrotic narrowing. CR has the potential to identify and even differentiate these forms of lymphoma due to the global 3D overview it provides and its capacity to accentuate texture differences between healthy tissue and disease processes. Furthermore, the presence of associated bulky lymphadenopathy and multifocal involvement can help differentiate lymphomas from small bowel GIST and adenocarcinoma.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Sarcomas only represent about 10% of small bowel cancers. Leiomyosarcomas are the most common, though many other subtypes can occur, each with distinct imaging features such as varying compositions . These tumors most frequently arise in the jejunum, followed by the ileum and duodenum and are typically slow growing yet aggressive . On imaging, they usually appear as large, heterogeneously enhancing masses with central necrosis and focal wall thickening and rarely tumoral calcification. Cavitation and direct communication with the bowel lumen may be observed, and their vascular nature often leads to complications such as bleeding or perforation.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - In summary, small bowel neoplasms are rare and often present with either nonspecific symptoms or acute emergencies, making diagnosis in the emergency setting pivotal. CECT remains the frontline imaging modality, but 3D CR offers added value by enhancing surface detail, accentuating subtle textural differences, and providing realistic depictions of tumor anatomy and surrounding structures. These advantages potentially improve detection, characterization, and assessment of vascular involvement, ultimately aiding radiologists in rapidly identifying and managing small bowel malignancies in acute care settings.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617
Chest
- Purpose: To quantify postimplementation concordance between a U.S. Food and Drug Administration–cleared artificial intelligence (AI) tool and AI-informed radiologists for pulmonary embolism (PE) detection at CT pulmonary angiography, with real-time adjudication of discordances.
Materials and Methods: A commercial PE AI tool was retrospectively implemented in the clinic across an integrated network (August 9, 2021 February 20, 2023). Adult CT pulmonary angiographic acquisitions underwent real-time AI analysis and radiologist interpretation. Radiologist-AI disagreements triggered adjudication by thoracic radiologists via the AI quality oversight process. Adjudicator diagnosis served as the reference standard for discordant cases. Concordance was measured and diagnostic performance of radiologists and AI was compared using adjudication for discordant cases.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017 - Results: A total of 32 501 CT pulmonary angiographic acquisitions obtained from 29 492 patients (mean age, 62.4 years Å} 18.6 [SD], 17 424 female patients) were evaluated. PE positivity was 9.93% (3226 of 32 501). Overall concordance was 97.79% (95% CI: 97.62, 97.94) and was higher for AI-negative than for AI-positive examinations (98.18% vs 93.75%; P < .001). Expert adjudication favored the radiologist in 88.73% of discordances. The rate of unique diagnosis by the interpreting radiologist (483 of 3226 [14.97%]) was approximately 19 times that of the AI tool alone (26 of 3226 [0.81%]). Concordance varied by PE features: acute versus chronic (87.34% vs 60.12%; P < .001) and location (central, 95.79%; lobar and/or segmental, 83.81%; subsegmental, 58.62%; all P < .001).
Conclusion: In large-scale deployment, AI showed high concordance with radiologists and made meaningful contributions in discordant reviews while expert oversight confirmed complementary roles and highlighted scenarios of radiologist-AI divergence.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017 - ■ In a retrospective study of 29 492 patients, radiologist–artificial intelligence (AI) concordance for pulmonary embolism (PE) detection was high (97.79%) and greater for AI-negative than for AI positive examinations (98.18% vs 93.75%), suggesting AI-negative outputs provide supportive signal while AI-positive alerts merit scrutiny.
■ Expert adjudication favored the radiologist in 88.73% of discordances,and AI contributed selectively, underscoring complementary roles in radiologist-led workflows.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017 - These findings support complementary roles for radiologists and AI, with radiologist-led oversight remaining essential. Radiologists should avoid overreliance on AI, remain vigilant in search patterns, and scrutinize AI-positive results. The clinical value of the AI tool is maximized under discerning, radiologist-led oversight, reinforcing AI’s role as diagnostic support and triage rather than a standalone solution. AI should be directly supervised by radiologists who render final diagnoses and should not be used in alternative workflows without imaging-expert oversight.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017 - In conclusion, we observed a high ~98% agreement between radiologists and the AI tool for the diagnosis of PE, with both AI and AI-informed radiologists offering unique diagnostic contributions. Future work will examine the attributes of radiologist- AI disagreement cases to determine which imaging, patient, or report factors are more likely to confound humanreaders compared with AI. We will also quantify the triage benefits of this tool, including its impact on turnaround time for PE-positive cases.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017
Deep Learning
- Purpose: To quantify postimplementation concordance between a U.S. Food and Drug Administration–cleared artificial intelligence (AI) tool and AI-informed radiologists for pulmonary embolism (PE) detection at CT pulmonary angiography, with real-time adjudication of discordances.
Materials and Methods: A commercial PE AI tool was retrospectively implemented in the clinic across an integrated network (August 9, 2021 February 20, 2023). Adult CT pulmonary angiographic acquisitions underwent real-time AI analysis and radiologist interpretation. Radiologist-AI disagreements triggered adjudication by thoracic radiologists via the AI quality oversight process. Adjudicator diagnosis served as the reference standard for discordant cases. Concordance was measured and diagnostic performance of radiologists and AI was compared using adjudication for discordant cases.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017 - Results: A total of 32 501 CT pulmonary angiographic acquisitions obtained from 29 492 patients (mean age, 62.4 years Å} 18.6 [SD], 17 424 female patients) were evaluated. PE positivity was 9.93% (3226 of 32 501). Overall concordance was 97.79% (95% CI: 97.62, 97.94) and was higher for AI-negative than for AI-positive examinations (98.18% vs 93.75%; P < .001). Expert adjudication favored the radiologist in 88.73% of discordances. The rate of unique diagnosis by the interpreting radiologist (483 of 3226 [14.97%]) was approximately 19 times that of the AI tool alone (26 of 3226 [0.81%]). Concordance varied by PE features: acute versus chronic (87.34% vs 60.12%; P < .001) and location (central, 95.79%; lobar and/or segmental, 83.81%; subsegmental, 58.62%; all P < .001). Conclusion: In large-scale deployment, AI showed high concordance with radiologists and made meaningful contributions in discordant reviews while expert oversight confirmed complementary roles and highlighted scenarios of radiologist-AI divergence.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017 - ■ In a retrospective study of 29 492 patients, radiologist–artificial intelligence (AI) concordance for pulmonary embolism (PE) detection was high (97.79%) and greater for AI-negative than for AI positive examinations (98.18% vs 93.75%), suggesting AI-negative outputs provide supportive signal while AI-positive alerts merit scrutiny.
■ Expert adjudication favored the radiologist in 88.73% of discordances,and AI contributed selectively, underscoring complementary roles in radiologist-led workflows.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017 - These findings support complementary roles for radiologists and AI, with radiologist-led oversight remaining essential. Radiologists should avoid overreliance on AI, remain vigilant in search patterns, and scrutinize AI-positive results. The clinical value of the AI tool is maximized under discerning, radiologist-led oversight, reinforcing AI’s role as diagnostic support and triage rather than a standalone solution. AI should be directly supervised by radiologists who render final diagnoses and should not be used in alternative workflows without imaging-expert oversight.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017 - In conclusion, we observed a high ~98% agreement between radiologists and the AI tool for the diagnosis of PE, with both AI and AI-informed radiologists offering unique diagnostic contributions. Future work will examine the attributes of radiologist- AI disagreement cases to determine which imaging, patient, or report factors are more likely to confound humanreaders compared with AI. We will also quantify the triage benefits of this tool, including its impact on turnaround time for PE-positive cases.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017
- Let us be brutally honest with ourselves: in many technical aspects, the machines will win. We must not dismiss the power of AI or pretend it is just another iterative tool. It is a formidable, paradigm-shifting technology. Soon, AI will seamlessly synthesize trillions of data points, catch invisible clinical patterns, and compute probabilities with a precision humans simply cannot match. Faced with this reality, the era of expecting physicians to act as human algorithms, as the benchmark interpreter of clinical data and evidence-based guidelines, is coming to an end.
Irreplaceable: Five Enduring Roles of the Physician in the AI Era.
Lin S.
Fam Med. 2026;58(5):397-398 - So how do we prepare for this new paradigm of doctoring? The future of our profession remains in our classrooms and clinics. To prepare our learners for the age of AI, we must fundamentally change how we teach. If we continue the status quo, we are training them to compete in a race they have already lost. Instead, our training must evolve to select for and cultivate empathy, resilience, deep listening, and moral courage. We must teach them how to be guides through complex lives, anchors in the storm, motivators of the weary, advocates against broken systems, and healers with their hands.
Irreplaceable: Five Enduring Roles of the Physician in the AI Era.
Lin S.
Fam Med. 2026;58(5):397-398 - In other words, every physician needs to be more like a family physician. My colleagues, we stand at the threshold of an identity defining transformation. For too long, we have been forced to act like machines—staring at screens, clicking boxes, and typing away our sanity. If machines are here to take that work back, I say we let them. The AI can do the computational heavy lifting, so we can do the relational heavy lifting. It can and should be a beautiful partnership. So do not fear AI. Engage it. Guide it. The future of medicine is not automated. The future is deeply, profoundly human. And in that future, you are irreplaceable.
Irreplaceable: Five Enduring Roles of the Physician in the AI Era.
Lin S.
Fam Med. 2026;58(5):397-398 - * THE REAL-WORLD GUIDE
* THE HUMAN ANCHOR
* THE MOTIVATOR
* THE SYSTEM ADVOCATE
* THE HANDS-ON HEALER
Irreplaceable: Five Enduring Roles of the Physician in the AI Era.
Lin S.
Fam Med. 2026;58(5):397-398
- IMPORTANCE Artificial intelligence (AI) systems for skin cancer detection perform well in controlled settings but frequently underperform in everyday clinical practice, raising critical questions about their readiness for deployment.
OBJECTIVE To compare the diagnostic accuracy of AI algorithms vs human evaluators across varying expertise levels for skin lesion diagnosis, including rare and atypical cases, in a realistic clinical context.
CONCLUSIONS AND RELEVANCE In this diagnostic study, a modern foundation model surpassed readers with less than 3 years of experience on accuracy of skin lesion diagnosis and matched those with 3 to 10 years of experience but remained inferior to experts with more than 10 years of experience, highlighting both the promise and current limitations of AI in dermatologic diagnosis.
Limits of Artificial Intelligence Models for Skin Cancer Diagnosis in Realistic Settings
Julien Anriot et al.
JAMA Dermatol. doi:10.1001/jamadermatol.2026.1492 - Question How does artificial intelligence (AI) diagnostic performance compare to human dermatologists of varying experience for skin cancer detection in realistic clinical settings? Findings In this diagnostic study of 652 physicians and 3 AI models evaluating 1117 cases, expert dermatologists (>10 years of experience) achieved the highest accuracy (74.2%), considerably outperforming a modern unimodal foundation model (72.2%), which exceeded dermatologists with less than 1 year of experience (59.1%), while the first-generation convolutional neural network underperformed allreaders(56.7%). Meaning Future practice should integrate human-AI collaboration, with AI supporting less experienced clinicians and providing expert triage assistance and help to minimize fatigue-related diagnostic errors.
Limits of Artificial Intelligence Models for Skin Cancer Diagnosis in Realistic Settings
Julien Anriot et al.
JAMA Dermatol. doi:10.1001/jamadermatol.2026.1492 - Results of this diagnostic study show that foundation models approached the diagnostic accuracy of well-trained clinicians and surpassed novices but still fell short of the best experts,who remain the reference standard. First-generation CNN models were no longer adequate for broad clinical testing, and the further development of foundation models will be essential to improve generalization across the full clinical spectrum. AI systems demonstrate strong potential as diagnostic support tools, particularly for early-career clinicians. Despite overconfident mainstream narratives about achieving clinical excellence through AI-based technology alone, it remains crucial to continue training primary care physicians to recognize skin lesions, especially cancers, and to continue educating dermatologists toward expertise. This training is important not only in dermoscopy, but also in the use of AI, including how it works and its limitations.
Limits of Artificial Intelligence Models for Skin Cancer Diagnosis in Realistic Settings
Julien Anriot et al.
JAMA Dermatol. doi:10.1001/jamadermatol.2026.1492 - “The underperformance of the multimodal system underscores that AI progress depends not only on data scale, but also on intelligent data integration. The future likely lies in collaboration between humans and machines to optimize diagnostic performance. For novice practitioners, AI could serve as a safety net and educational tool. For experts, it could provide an efficient triage modality and a systematic second reading, particularly useful for reducing errors caused by fatigue or inattention.”
Limits of Artificial Intelligence Models for Skin Cancer Diagnosis in Realistic Settings
Julien Anriot et al.
JAMA Dermatol. doi:10.1001/jamadermatol.2026.1492
- An agentic system, in contrast, may operate like a full - fledged clinician charged with caring for the patient as a whole. Having defined shared health goals with the patient, it could be tasked with reducing their 10 -year atherosclerotic cardiovascular disease risk score. It might calculate current risk, order follow -up laboratory tests, ensure that the patient was prescribed and is taking statins, recommend a smoking cessation program, and enroll them in supplemental employee insurance that covers nicotine patches.
A Typology of Generative Health Care Artificial Intelligence — Definitions and Policy Implications
David Blumenthal , Vivian S. Lee
NEJM AI 2026;3(6) - Agentic AI systems can operate at scale. Programmed at the level of a chief medical officer, such a system could be designed to achieve target blood pressure control across a health system’s patient population, thereby optimizing value -based reimbursement. If granted the autonomy and access, it might independently decide to scan all electronic health records, identify patients with blood pressures that exceed target levels, and prompt clinicians to take patient-specific steps. With clinician agreement, it could prescribe medications, schedule tests and appointments, and inform patients, repeating this process — under human supervision — until population -level goals are achieved.
A Typology of Generative Health Care Artificial Intelligence — Definitions and Policy Implications
David Blumenthal , Vivian S. Lee
NEJM AI 2026;3(6) - AI agents pose different policy and management challenges. As agents produce actions as well as information, they offer both greater risks and greater benefits than the LLMs on which they are built. An agent that is trained and marketed to accomplish a specific clinical task, such as minimizing cardiovascular risk, will likely be classified by the U.S. Food and Drug Administration (FDA) as Software as a Medical Device (SaMD) and be subject to FDA regulation. If the FDA were not to invoke this authority, it would likely face legal challenges demanding that it enforce current law.
A Typology of Generative Health Care Artificial Intelligence — Definitions and Policy Implications
David Blumenthal , Vivian S. Lee
NEJM AI 2026;3(6) - Agentic systems raise additional issues that require alternative channels for review and oversight. One is that they may be trained to accomplish diverse tasks, such as comprehensive care for patients or complex organizational goals. In the former case, they may pose unprecedented regulatory challenges for the FDA because of the technical challenges in conducting premarket assessment of devices that undertake a wide range of clinical activities and that can change and learn over time. Novel regulatory or oversight pathways may be needed to deal with the capabilities of agentic systems.
A Typology of Generative Health Care Artificial Intelligence — Definitions and Policy Implications
David Blumenthal , Vivian S. Lee
NEJM AI 2026;3(6) - The distinctions among LLMs, agents, and agentic systems matter because their inherent properties, capabilities, and uses raise different quality and safety issues — and different challenges for oversight. Some forms may be subject to minimal direct oversight. For others, oversight may vary with use case. Relevant factors include clinical versus nonclinical, level of direct patient risk, and existing regulatory authorities. As they are used more widely, increased scrutiny of the underlying values embedded in these models — values that influence their recommendations and behaviors — may prompt further ethical review. In addition, agent -to -agent interactions will create new challenges as patient, provider, pharmaceutical, and insurer agents interact across administrative matters and clinical care.
A Typology of Generative Health Care Artificial Intelligence — Definitions and Policy Implications
David Blumenthal , Vivian S. Lee
NEJM AI 2026;3(6) - Understanding—or interpretability—of AI need not mean grasping every line of code or every neural-network parameter. Just as we study human behavior at multiple levels, from neuroscience to psychology to sociology, AI principles and operations can be explored and understood at varying levels. Full mechanistic understanding may remain elusive, but science is rarely all-or-nothing; partial understanding is still useful. What makes such understanding urgent is not a demand for completeness but a practical need: As capabilities accelerate, even imperfect causal insights into AI systems may let us detect risks early and intervene before harm compounds.
A narrowing window to understand AI
Eric Horvitz and Robert West
Science 392 (6802), June 2026. DOI: 10.1126/science.aei3167 - One trend challenging understanding is the rise of AI-directed AI design. AI systems are now designed and refined by AI systems through recursive cycles that can outpace human understanding and unfold in high-dimensional spaces that resist intuition. The result isgrowing operational opacity: Performance improves, while insight into how it is achieved diminishes. To promote human insight and control, AI systems that contribute to their own design should produce explanations and tools that make their architecture and operation intelligible to humans. Otherwise, opacity may emerge as an unintended consequence of the design process itself.
A narrowing window to understand AI
Eric Horvitz and Robert West
Science 392 (6802), June 2026. DOI: 10.1126/science.aei3167 - More subtle is the possibility that we will gradually lose interest in understanding and guiding AI. As AI systems become deeply embedded in human environments, they may respond to preferences but also shape them. Systems optimized for engagement or approval may reduce friction and discourage scrutiny. Over time, curiosity and skepticism may erode, leading to neglect and acceptance. Preserving human agency must therefore remain a central goal. It is not enough to monitor how AI systems behave. We must also understand how they shape human goals and judgment, and ensure that people retain the capacity and motivation to question, audit, and guide them.
A narrowing window to understand AI
Eric Horvitz and Robert West
Science 392 (6802), June 2026. DOI: 10.1126/science.aei3167 - The goal is not just more capable AI, but AI that is more intelligible, accountable, and aligned with human aims. The window for achieving that future is narrowing. Without sustained efforts to keep AI intelligible, we may come to depend on systems that we can neither adequately understand nor effectively guide—transforming the relationship between people and the systems they create.
A narrowing window to understand AI
Eric Horvitz and Robert West
Science 392 (6802), June 2026. DOI: 10.1126/science.aei3167
- “To paraphrase ChatGPT -5.1, an LLM is a computer program that is trained on huge amounts of existing text, learns language patterns, and then generates text, answers questions, and summarizes documents. Capabilities also extend to translating between different languages and dialects, as well as converting conversational language into computer code and back, a skill often described as vibe coding. ”
A Typology of Generative Health Care Artificial Intelligence — Definitions and Policy Implications
David Blumenthal , Vivian S. Lee
NEJM AI 2026;3(6) - As these models have expanded beyond language -based tasks, they have come to be called “foundation models.” In health care, this means they can handle diverse types of data such as images, audio, and time -series data (e.g., from an electrocardiogram or continuous glucose monitor tracing). LLMs generate responses probabilistically, predicting the likely next word rather than following fixed rules, which introduces a degree of variability that can make responses appear more natural or creative. However, in clinical settings, where accuracy and consistency are critical, such variability can present challenges.3
A Typology of Generative Health Care Artificial Intelligence — Definitions and Policy Implications
David Blumenthal , Vivian S. Lee
NEJM AI 2026;3(6) - Despite their remarkable abilities, stand -alone LLMs have important limitations. First, they rely solely on their training data; they do not spontaneously acquire new information. Second, LLMs have no memory — they do not retain their past experience, previous answers to questions, or patient information from one task to the next. Third, their output depends heavily on the ways in which they are instructed. Fourth, they produce language or data output but do not act on their responses. Fifth, they do not independently set or fulfill goals but must be directed through a prompt to produce a product.
A Typology of Generative Health Care Artificial Intelligence — Definitions and Policy Implications
David Blumenthal , Vivian S. Lee
NEJM AI 2026;3(6)
- IMPORTANCE Artificial intelligence (AI) is changing health and health care on an unprecedented scale. Though the potential benefits are massive, so are the risks. The JAMA Summit on AI discussed how health and health care AI should be developed, evaluated, regulated, disseminated, and monitored.
CONCLUSIONS AND RELEVANCE AI will disrupt every part of health and health care delivery in the coming years. Given the many long-standing problems in health care, this disruption represents an incredible opportunity. However, the odds that this disruption will improve health for all will depend heavily on the creation of an ecosystem capable of rapid, efficient, robust, and generalizable knowledge about the consequences of these tools on health.
AI, Health, and Health Care Today and Tomorrow
The JAMA Summit Report on Artificial Intelligence
Derek C. Angus et al.
JAMA. 2025;334(18):1650-1664. - Ensuring AI is deployed equitably and in a manner that improves health outcomes or, if improving efficiency of health care delivery, does so safely, requires progress in 4 areas. First, multistakeholder engagement throughout the total product life cycle is needed. This effort would include greater partnership of end users with developers in initial tool creation and greater partnership of developers, regulators, and health care systems in the evaluation of tools as they are deployed. Second, measurement tools for evaluation and monitoring should be developed and disseminated. Beyond proposed monitoring and certification initiatives, this will require new methods and expertise to allow health care systems to conduct or participate in rapid, efficient, and robust evaluations of effectiveness. The third priority is creation of a nationally representative data infrastructure and learning environment to support the generation of generalizable knowledge about health effects of AI tools across different settings. Fourth, an incentive structure should be promoted, using market forces and policy levers, to drive these changes.
AI, Health, and Health Care Today and Tomorrow
The JAMA Summit Report on Artificial Intelligence
Derek C. Angus et al.
JAMA. 2025;334(18):1650-1664. - The following 3 advances help differentiate AI from prior digital technologies:
• Deep learning: development of deeper, more convoluted neural networks capable of interpreting large complex datasets to address specific yet complicated tasks (eg, computer vision).
• Generative AI: an extension of deep learning using so-called large language and foundation models capable of generating new content to address far broader task requests (eg, ChatGPT or Gemini).
• Agentic AI: an extension of deep learning and generative A capable of autonomous decision-making (eg, the Tesla autopilotsoftware for autonomous driving).
AI, Health, and Health Care Today and Tomorrow
The JAMA Summit Report on Artificial Intelligence
Derek C. Angus et al.
JAMA. 2025;334(18):1650-1664. - Health and health care AI tools should be subject to a governance structure that protects individuals and ensures the tools achieve their potential benefits. For other health care interventions, regulatory oversight is an important part of that governance, assuring society and markets that an intervention is credible. However, the US has no comprehensive fit-for-purpose regulatory framework for health and health care AI. Reasons include the diverse and rapidly evolving nature of AI technology, the numerous agencies with jurisdiction over different types and aspects of AI, and a lack of regulatory frameworks specifically tailored forAI within these agencies.41 Drug and traditional medical device development also benefits from international harmonization of regulatory standards.
- Finally, use of AI tools has thus far largely been voluntary. However, as their benefits become more established, failure to useanAItoolmaybeconsideredunethicalorabreachofstandardcare.Ahealth care system or professional may thus be liable in a malpractice suit for failing to use AI. At the same time, if a plaintiff sues for an adverse outcome when care was provided in which an AI tool was involved, the question arises of whether liability rests with the health care professional, the health care system,or the developer of the tool. Though relevant case law is currently limited, developers, health care systems, and health care professionals will all have to adopt strategies to manage their liability risk. These examples are just some of the many new issues that will need to be addressed as Aibecomes more incorporated in health and health care.
AI, Health, and Health Care Today and Tomorrow
The JAMA Summit Report on Artificial Intelligence
Derek C. Angus et al.
JAMA. 2025;334(18):1650-1664. - AI will massively disrupt health and health care delivery in the coming years. The traditional approaches to evaluate, regulate,and monitor novel health care interventions are being pushed to their limits, especially with generative and agentic AI, and especially because the tools’ effects cannot be fully understood until deployed in practice. Nonetheless,many tools are already being rapidly adopted,in part because they are addressing important pain points for end users. Given the many long-standing problems in health care, this disruption represents an incredible opportunity. However, the odds that this disruption will improve health for all will depend heavily on creation of an ecosystem capable of rapid, efficient, robust, and generalizable knowledge about the consequences of these tools on health.
AI, Health, and Health Care Today and Tomorrow
The JAMA Summit Report on Artificial Intelligence
Derek C. Angus et al.
JAMA. 2025;334(18):1650-1664. - Pancreatic cystic lesions (PCLs) are increasingly detected due to the widespread use of cross-sectional imaging and represent a significant diagnostic challenge because of their heterogeneous biological behavior, ranging from benign lesions to neoplasms with malignant potential. Accurate characterization and risk stratification are essential to guide appropriate management and avoid unnecessary surgical interventions. Conventional imaging modalities, including computed tomography (CT), magnetic resonance (MR) imaging, and endoscopic ultrasound (EUS), remain central to the diagnostic work-up; however, their ability to reliably differentiate cyst subtypes and predict malignant transformation remains limited. In recent years, artificial intelligence (AI) and radiomics have emerged as promising approaches for improving the non-invasive characterization of PCLs by extracting quantitative imaging features beyond those appreciable through visual assessment.
Radiomics and artificial intelligence in pancreatic cyst characterization: future or fiction?
Cesare Maino · Paolo Niccolò Franco · Federica Omboni
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-026-05515-z - This narrative review summarizes the current evidence regarding CT- and MR-based radiomics and AI in pancreatic cyst characterization, focusing on their role in differentiating mucinous from non-mucinous cysts, identifying high-risk intraductal papillary mucinous neoplasms (IPMNs), and supporting clinical decision-making. The potential advantages of these techniques are discussed alongside main methodological limitations, including variability in imaging acquisition protocols, segmentation reproducibility, small and often retrospective datasets, limited external validation, and interpretability of AI-based models. Further multicenter studies, standardized radiomic pipelines, and prospective validation are required before these tools can be reliably integrated into routine clinical practice.
Radiomics and artificial intelligence in pancreatic cyst characterization: future or fiction?
Cesare Maino · Paolo Niccolò Franco · Federica Omboni
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-026-05515-z - CT is often the first modality in which PCLs are incidentally detected, with reported accuracy of approximately 60–75% in distinguishing cysts with or without malignant potential (highest reported sensitivity and specificity of 72% and 74%, respectively). While CT offers high spatial resolution for size measurement of medium-large cysts, it underperforms for small lesion characterization . Advances such as spectral and photon-counting CT may partly address this limitation; however, CT remains inferior to MR and endoscopic ultrasound (EUS) in detecting septa and cyst with main pancreatic duct (MPD) communication.
Radiomics and artificial intelligence in pancreatic cyst characterization: future or fiction?
Cesare Maino · Paolo Niccolò Franco · Federica Omboni
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-026-05515-z - Accurate and reproducible segmentation of the region of interest (ROI) is fundamental to reliable radiomic feature extraction. In PCLs, segmentation is particularly challenging due to irregular lesion morphology, partial volume effects, ill-defined cyst walls, and the presence of septa or mural nodules . Manual segmentation by expert radiologists remains the reference standard but is time-consuming and subject to significant inter- and intra-observer variability. Semi-automated and DL-based segmentation approaches can improve efficiency and reproducibility, though their performance depends heavily on image quality, lesion heterogeneity, and annotated training data availability.
Radiomics and artificial intelligence in pancreatic cyst characterization: future or fiction?
Cesare Maino · Paolo Niccolò Franco · Federica Omboni
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-026-05515-z - Worrisome features include cyst size greater than 3 cm, enhancing mural nodule less than 5 mm, thickened or enhancing walls, MPD dilatation 5 to 9 mm, abrupt change in caliber of pancreatic duct with distal atrophy, lymphadenopathy, increase in serum CA 19-9, cyst growth rate greater than mm/2 years.
Pancreatic Cancer Screening and Early Detection
Current Strategies and Emerging Innovations in Imaging
Shravya Srinivas Rao , Ok Kyu Song, Yoshifumi Noda et al.
Radiol Clin N Am 64 (2026) 489–499 - A multicenter AI tool on CT scans (PANDA model) reached a sensitivity of 93% and specificity of 99% highlighting its potential in large-scale PDAC screening. 36 Radiomics, which extracts quantitative imaging features invisible to human readers, further boosts early detection. Radiomics-based machine learning classifiers can distinguish early cancer from normal tissue with high accuracy. 35 These models can capture subtle textural changes years before the tumor is evident, enabling identification of high-risk individuals even outside known familial risk groups. Pancreatic Cancer Screening and Early Detection
Current Strategies and Emerging Innovations in Imaging
Shravya Srinivas Rao , Ok Kyu Song, Yoshifumi Noda et al.
Radiol Clin N Am 64 (2026) 489–499 - • Individuals with strong family history or pathogenic germline mutations (eg, breast cancer gene 2 [BRCA2], CDKN2A, serine/threonine kinase 11 [STK11], and serine protease 1 [PRSS1]) have greater than 5% lifetime PDAC risk and should undergo pancreatic cancer screening by annual MRI/MRCP ± EUS surveillance at specialized centers.
• Pancreatic cystic lesions such as MCNs and IPMNs are precursors of pancreatic cancer, necessitating tailored risk-based surveillance strategies.
Pancreatic Cancer Screening and Early Detection
Current Strategies and Emerging Innovations in Imaging
Shravya Srinivas Rao , Ok Kyu Song, Yoshifumi Noda et al.
Radiol Clin N Am 64 (2026) 489–499 - “The pancreatic cancer diagnosis: radiologists meet AI (PANORAMA) study showed that AI models outperformed 68 radiologists (AUC = 0.92 vs. 0.88) across multi-institutional datasets, underscoring robust performance across different settings. Another study by Chen et al. validated a detection model in a nationwide cohort (n = 1,473), reporting an AUC of 0.95 and sensitivity of 89.7%, with preserved performance for tumors <2 cm in diameter. Together, these findings suggest the potential use of AI in detecting pancreatic tumors early.”
From Bench to Bedside: The Path Toward Real-World Translation for Artificial Intelligence in Pancreatic Cancer Detection.
Syailendra EA, Arshad H, Lopez-Ramirez F, Tixier F, Kawamoto S, Fishman EK, Chu LC.
Korean J Radiol. 2026 Jun;27(6):543-554. doi: 10.3348/kjr.2026.0003. PMID: 42225574; PMCID: PMC13236442. - Retrospective studies showed that progressive pre-diagnostic pancreatic changes such as parenchymal atrophy, increased tissue heterogeneity, and ductal dilation preceded the diagnosis of PDAC . Several groups have evaluated whether AI can retrospectively detect PDAC on pre-diagnostic CT (i.e., obtained before the clinical diagnosis of PDAC). Mukherjee et al. analyzed scans obtained up to 3 years before diagnosis (median lead time, 398 days) and developed radiomics-based classifiers that achieved an AUC of 0.98, demonstrating measurable pancreatic texture and shape differences well before visible tumors formed. Using a CNN model, Korfiatis et al. demonstrated promising results (AUC = 0.97) for pre-diagnostic detection, with a median lead time of 475 days.
From Bench to Bedside: The Path Toward Real-World Translation for Artificial Intelligence in Pancreatic Cancer Detection.
Syailendra EA, Arshad H, Lopez-Ramirez F, Tixier F, Kawamoto S, Fishman EK, Chu LC.
Korean J Radiol. 2026 Jun;27(6):543-554. doi: 10.3348/kjr.2026.0003. PMID: 42225574; PMCID: PMC13236442. - Clinician acceptance remains vital for effective use of AI models. Limited interpretability and the black-box nature of the models in generating predictions reduce clinicians’ willingness to rely on AI. Although explainability tools may enhance transparency, their impact on diagnostic performance is variable. Early involvement of radiologists in task definition, error tolerance, and interface design has been shown to improve clinical relevance and usability, supporting higher adoption rates [82,83]. Patient trust is also a critical factor. Surveys show that confidence in AI for diagnosis or treatment is modest, with many patients expressing discomfort and low trust in health systems to use AI responsibly [84,85]. Addressing these issues requires co-design with end users, task explainability, and communication strategies that clearly articulate the role of AI as an assistive and not autonomous component of the diagnostic process.
From Bench to Bedside: The Path Toward Real-World Translation for Artificial Intelligence in Pancreatic Cancer Detection.
Syailendra EA, Arshad H, Lopez-Ramirez F, Tixier F, Kawamoto S, Fishman EK, Chu LC.
Korean J Radiol. 2026 Jun;27(6):543-554. doi: 10.3348/kjr.2026.0003. PMID: 42225574; PMCID: PMC13236442. - Early detection using AI models provides an opportunity to improve the outcomes of pancreatic cancer patients. Multiple studies showed promising results; however, these remain limited to research settings. Progress will require efforts across different aspects, as adoption depends on factors beyond model specifics, including workflow integration, governance, financial commitment, reimbursement, and ultimately, provider and patient trust. Equally important is post-deployment surveillance to monitor model behaviors clinically. The path forward will depend on AI models that not only demonstrate exceptional performance but also operate reliably within the complex and evolving clinical and societal environments that shape pancreatic cancer care.
From Bench to Bedside: The Path Toward Real-World Translation for Artificial Intelligence in Pancreatic Cancer Detection.
Syailendra EA, Arshad H, Lopez-Ramirez F, Tixier F, Kawamoto S, Fishman EK, Chu LC.
Korean J Radiol. 2026 Jun;27(6):543-554. doi: 10.3348/kjr.2026.0003. PMID: 42225574; PMCID: PMC13236442.
- Purpose: The purpose of this study was to evaluate the contribution of radiomics features extracted from various pancreatic structures on computed tomography (CT) images, including the main pancreatic duct and cystic lesion, for predicting the pathological grade of intraductal papillary mucinous neoplasms (IPMNs) using machine learning models Conclusion: Integrating CT-based radiomics features from pancreatic ducts and cysts improves classification performance, with main pancreatic duct features being the most contributive predictor of IPMN grade.
Evaluating the role of the main pancreatic duct in intraductal papillary mucinous neoplasm grading: A multi-structure radiomics-based machine learning approach.
Syailendra EA, Lopez-Ramirez F, Tixier F, Blanco A, Arshad H, Yasrab M, Rahmatullah ZF, Javed AA, He J, Lennon AM, Hruban RH, Afghani E, Kawamoto S, Chu LC, Fishman EK.
Diagn Interv Imaging. 2026 Feb (in press) - The purpose of this study was to evaluate the contribution of radiomics features extracted from different pancreatic structures, including the main pancreatic duct (MPD) and cystic lesion, for predicting the pathological grade of IPMNs using machine learning models.
Evaluating the role of the main pancreatic duct in intraductal papillary mucinous neoplasm grading: A multi-structure radiomics-based machine learning approach.
Syailendra EA, Lopez-Ramirez F, Tixier F, Blanco A, Arshad H, Yasrab M, Rahmatullah ZF, Javed AA, He J, Lennon AM, Hruban RH, Afghani E, Kawamoto S, Chu LC, Fishman EK.
Diagn Interv Imaging. 2026 Feb (in press) 
- Although not significant, the cyst radiomics model outperformed the MPD radiomics model. The improved performance after incorporating MPD features aligns with the groundwork of pancreatic ductal changes in IPMN pathogenesis. High-grade IPMNs are associated with more extensive MPD involvement, inflammatory changes, and architectural distortion that may be captured through quantitative texture analysis.
Evaluating the role of the main pancreatic duct in intraductal papillary mucinous neoplasm grading: A multi-structure radiomics-based machine learning approach.
Syailendra EA, Lopez-Ramirez F, Tixier F, Blanco A, Arshad H, Yasrab M, Rahmatullah ZF, Javed AA, He J, Lennon AM, Hruban RH, Afghani E, Kawamoto S, Chu LC, Fishman EK.
Diagn Interv Imaging. 2026 Feb (in press) - Previous studies have explored the use of radiomics for predicting IPMN grades with variable performance. However, most prior investigations focused on either cyst-derived or pancreas-derived features [20,21,29,32]. We systematically evaluated the individual and combined contributions of radiomics features derived from both the IPMNs lesion and MPD, allowing for a comprehensive understanding of structure-specific information. From the failure analysis, FPs were driven by MPD-derived features, with SHAP analysis showing greater MPD contribution compared to cyst-based texture features that characterized TNs. We found that FPs clustered closer to TPs and statistical analysis showing no difference between FPs and TPs in some radiomics features. This may suggest that errors arise from radiomics similarities between benign lesions and true high-grade lesions, highlighting the limits of imaging-only approaches and the potential role of multimodal integration to improve specificity.
Evaluating the role of the main pancreatic duct in intraductal papillary mucinous neoplasm grading: A multi-structure radiomics-based machine learning approach.
Syailendra EA, Lopez-Ramirez F, Tixier F, Blanco A, Arshad H, Yasrab M, Rahmatullah ZF, Javed AA, He J, Lennon AM, Hruban RH, Afghani E, Kawamoto S, Chu LC, Fishman EK.
Diagn Interv Imaging. 2026 Feb (in press) - Several research directions should emerge from this work. First, development of automated segmentation algorithms for pancreatic structures would enable broader clinical implementation by eliminating the need for manual segmentation. Towards this goal, deep learning approaches show promise for accurate pancreatic anatomy delineation. Second, ensemble modeling approaches combining multiple structure- specific models based on segmentation availability could maximize clinical utility. For instances where MPD segmentation is challenging, cyst-based models could provide fallback predictions. Finally, multi- institutional collaborations would enable larger dataset and external validation to address generalizability concerns, such as the inclusion of other types of pancreatic cyst.
Evaluating the role of the main pancreatic duct in intraductal papillary mucinous neoplasm grading: A multi-structure radiomics-based machine learning approach.
Syailendra EA, Lopez-Ramirez F, Tixier F, Blanco A, Arshad H, Yasrab M, Rahmatullah ZF, Javed AA, He J, Lennon AM, Hruban RH, Afghani E, Kawamoto S, Chu LC, Fishman EK.
Diagn Interv Imaging. 2026 Feb (in press) - In conclusion, this study demonstrated that multi-structure radiomics improved the performance of pre-surgical prediction of IPMN grade compared to morphological feature such as diameter. MPD radiomics provides valuable discriminative information, complementing cyst-based features and contributing to model performance. Further validation and prospective clinical studies are needed to establish the real-world impact.
Evaluating the role of the main pancreatic duct in intraductal papillary mucinous neoplasm grading: A multi-structure radiomics-based machine learning approach.
Syailendra EA, Lopez-Ramirez F, Tixier F, Blanco A, Arshad H, Yasrab M, Rahmatullah ZF, Javed AA, He J, Lennon AM, Hruban RH, Afghani E, Kawamoto S, Chu LC, Fishman EK.
Diagn Interv Imaging. 2026 Feb (in press) - Purpose To evaluate the value of a combined model based on multiphase contrast-enhanced CT radiomics and clinical features for differentiating hypovascular pancreatic neuroendocrine tumors (hypo-PNETs) from pancreatic ductal adenocarcinoma (PDAC).
Conclusion A combined model integrating multiphasic CT radiomics and clinical features showed promising performance for differentiating hypo-PNETs from PDAC. This model may provide complementary support for preoperative diagnosis, although its incremental value over the radiomics-only model requires further validation.
Radiomics based on multiphasic contrast-enhanced CT combined with clinical features for differentiating hypovascular pancreatic neuroendocrine tumors and pancreatic ductal adenocarcinoma
Fan Xia · Jie Yu · Jianhua Wang · Zhongqiu Wang
J Cancer Res Clin Oncol. 2026 Apr 25;152(4):96. - Pancreatic duct dilatation, tumor composition, age, and maximum tumor diameter were identified as independent predictors. Among the radiomics models, the combined three-phase SVM model achieved the best performance, with AUCs of 0.814 and 0.812 in the training and test sets, respectively. The combined model yielded the highest AUCs (0.886 in the training set and 0.849 in the test set); however, because several between-model comparisons in the test set did not reach statistical significance, its advantage should be interpreted as a potential incremental benefit rather than definitive superiority.
Radiomics based on multiphasic contrast-enhanced CT combined with clinical features for differentiating hypovascular pancreatic neuroendocrine tumors and pancreatic ductal adenocarcinoma
Fan Xia · Jie Yu · Jianhua Wang · Zhongqiu Wang
J Cancer Res Clin Oncol. 2026 Apr 25;152(4):96. - From a translational perspective, the combined model may serve as a decision-support tool rather than a standalone diagnostic system. In future work, it could be converted into a nomogram or a web-based calculator to facilitate clinical application. Because contrast-enhanced CT is routinely available, such a model may be particularly useful in centers with limited expertise in pancreatic tumor imaging, where it could provide complementary information for radiologists and multidisciplinary teams when conventional imaging findings overlap.
Radiomics based on multiphasic contrast-enhanced CT combined with clinical features for differentiating hypovascular pancreatic neuroendocrine tumors and pancreatic ductal adenocarcinoma
Fan Xia · Jie Yu · Jianhua Wang · Zhongqiu Wang
J Cancer Res Clin Oncol. 2026 Apr 25;152(4):96.
GU Misc
- Germ Cell Testicular Tumor
Seminomas (approx. 50%): Tend to appear as more homogeneous, well-circumscribed, and uniformly enhancing soft-tissue masses.
Non-Seminomatous Germ Cell Tumors (NSGCTs - approx. 50%): These include embryonal carcinoma, yolk sac tumor, choriocarcinoma, and teratoma. On CT, they typically present as highly heterogeneous, ill-defined masses with cystic/low-attenuation components representing central necrosis, hemorrhage, or fluid-filled spaces (especially common in teratomas). Foci of calcification are also frequently seen. - Nodal Staging: Retroperitoneal
Right testis: Drains primarily to the interaortocaval, precaval, and pre-aortic nodes.
Left testis: Drains primarily to the para-aortic and pre-aortic nodes (near the left renal vein hilum).
Note: Pelvic or inguinal lymph node involvement is rare and typically only occurs if there has been prior scrotal or inguinal surgery (altering normal lymphatic pathways) or direct tumor invasion into the scrotal wall. - Testicular Tumor: Metastases
Lungs: The most common site of visceral hematogenous metastasis.
NSGCTs frequently present as multiple, small, bilateral, peripheral pulmonary nodules.
Seminomas tend to present as fewer, larger, more homogeneous masses. - Testicular Tumor: Metastases
Mediastinum & Supraclavicular Nodes: Seminomas characteristically spread in a contiguous manner up the thoracic duct, often causing low-attenuation masses in the posterior mediastinum. NSGCTs tend to spread less predictably, frequently involving the anterior mediastinum, hila, and supraclavicular/cervical nodes. - Testicular Tumor: Metastases
Liver: The second most common visceral site. Metastases appear hypodense on contrast-enhanced CT. Choriocarcinoma metastases are notoriously hypervascular and hemorrhagic, sometimes appearing hyperdense on non-contrast CT.
Brain and Bone: Rare at initial presentation, but checked via CT/MRI if neurological symptoms or bone pain are present.
Kidney
- The TNC acquisition is the established reference standard for detecting urinary stones. It informs clinical management by providing precise data on stone size and location, which are the primary factors used to predict spontaneous passage. Stones are frequently obscured by contrast media in the nephrographic and excretory phases, making the TNC acquisition essential for their detection. Alternative techniques, such as VNC imaging, have shown reduced sensitivity in the detection of small (< 3–5 mm)or low-attenuation stones. A meta-analysis of 13 studies found that VNC images have a pooled sensitivity for detecting urinary stones of only 78.1%, with individual study sensitivities ranging from 53% to 95%.
Noncontrast Acquisition in CT Urography Protocols: Counterpoint—Why It Remains Essential
André Euler
AJR 2026; 226:e2533641 - This limitation introduces two clinical problems. First, the lack of a definitive diagnosis in patients presenting with, for example, acute flank pain or hematuria, can lead to patient anxiety and further downstream testing. Second, there is the potential for underestimation of the clinical significance of asymptomatic stones 5 mm and smaller, which have been shown to require surgical treatment within 5 years in approximately 20% of patients. Improved VNC and novel virtual non-iodine algorithms from photon-counting CT (PCCT) are promising tools to overcome this limitation. However, recent literature suggests that a substantial percentage (up to 16%) of small stones are still erroneously subtracted using these novel approaches, limiting their current clinical applicability.
Noncontrast Acquisition in CT Urography Protocols: Counterpoint—Why It Remains Essential
André Euler
AJR 2026; 226:e2533641 - The TNC acquisition provides an essential quantitative reference standard for defining lesion enhancement. An increase in attenuation of 20 HU or greater from the TNC images to the contrast-enhanced phase is generally considered unequivocal enhancement. This metric is critical for differentiating solid renal masses from complex proteinaceous or hemorrhagic cysts.
Noncontrast Acquisition in CT Urography Protocols: Counterpoint—Why It Remains Essential
André Euler
AJR 2026; 226:e2533641 - The TNC acquisition acts as a crucial problem solver, assisting in establishing a definitive diagnosis and identifying potential mimickers, thereby minimizing patient anxiety. In an era of risk-stratified imaging, in which every CT urography examination should provide maximum diagnostic value, the harm of a missed, delayed, or false-positive diagnosis decisively outweighs the low radiation burden of a noncontrast scan. Therefore, TNC imaging remains an indispensable component of CT urography protocols.
Noncontrast Acquisition in CT Urography Protocols: Counterpoint—Why It Remains Essential
André Euler
AJR 2026; 226:e2533641 - In conclusion, with proper protocol optimization, VNC imaging can replace TNC acquisitions in CT urography without compromising diagnostic performance. VNC images have sensitivity comparable with that of TNC images for detecting stones 5 mm and larger, particularly when protocols incorporate oral hydration,iodinated contrast media dose optimization, and thin slices. For renal mass evaluation, VNC images offer superior anatomic registration and sufficient accuracy in enhancement assessment despite minor differences in attenuation when compared with TNC images. As DECT and PCCT become more widespread, VNC imaging represents a feasible strategy to streamline CT urography protocols, reduce radiation exposure, and maintain high diagnostic quality.
Noncontrast Acquisition in CT Urography Protocols: Point—No Longer Necessary in the Age of Dual-Energy and Photon-Counting CT
Ryan Chung, Avinash R. Kambadakone
AJR 2026; 226:e2533988
- The global burden of renal cell carcinoma (RCC) has risen substantially over the past three decades, while mortality rates have remained largely stable. This epidemiologic paradox suggests that a significant proportion of detected renal tumors may reflect overdiagnosis. This review synthesizes current evidence on the magnitude, drivers, and implications of overdiagnosis in RCC, integrating epidemiologic trends, tumor biology, imaging practices, and translational advances. The widespread use of cross-sectional imaging, particularly computed tomography (CT) and magnetic resonance imaging (MRI), has markedly increased the incidental detection of RCCs, with more than 50% now diagnosed incidentally. Many of these small renal masses (SRMs) are benign or biologically indolent. While early detection may benefit patients with aggressive disease, it can also lead to overtreatment, psychological burden, and increased healthcare costs. Emerging strategies, such as active surveillance, radiomics, and artificial intelligence (AI)-based risk stratification, offer potential pathways to reduce harm. Addressing overdiagnosis is essential to advancing a more precise and clinically meaningful approach to RCC detection and management.
Overdiagnosis in renal cancer: the hidden consequence of modern imaging
Katherine Vallecilla Valencia . Herney Andres Garcia‑Perdomo
International Urology and Nephrology https://doi.org/10.1007/s11255-026-05173-6 - While this shift represents progress in diagnostic capabilities, it also introduces a paradox: the incidence of RCC has nearly doubled in several regions over the past 30 years, yet mortality rates have remained stable or even declined slightly. This disparity suggests that some detected tumors may be clinically insignificant, reflecting the phenomenon of overdiagnosis: the identification of disease that would not have become symptomatic or life-threatening during a patient’s lifetime.
Overdiagnosis in renal cancer: the hidden consequence of modern imaging
Katherine Vallecilla Valencia . Herney Andres Garcia‑Perdomo
International Urology and Nephrology https://doi.org/10.1007/s11255-026-05173-6 - In this context, it is important to distinguish related but conceptually distinct terms. Overdetection is the increased identification of abnormalities through more sensitive or widespread imaging, including lesions of uncertain clinical relevance. Overdiagnosis is a subset of overdetection and specifically denotes the diagnosis of indolent disease that would not affect patient outcomes. In contrast, overtreatment is the unnecessary treatment of such lesions, exposing patients to potential harm without clinical benefit.
Overdiagnosis in renal cancer: the hidden consequence of modern imaging
Katherine Vallecilla Valencia . Herney Andres Garcia‑Perdomo
International Urology and Nephrology https://doi.org/10.1007/s11255-026-05173-6 - Active surveillance data demonstrate that small renal masses typically grow slowly, with median growth rates of approximately 0.2–0.3 cm per year and metastatic progression rates of less than 2% during follow-up. The indolent behavior of many incidentally detected lesions is consistent with findings from renal tumor biopsy series, including those of Richard et al., which support the notion that a substantial proportion of SRMs exhibit low-grade histology and slow progression. Many lesions remain stable for years without significant change, suggesting that routine surgical excision of all SRMs may constitute overtreatment. However, these findings should be interpreted cautiously, as patient selection remains critical.
Overdiagnosis in renal cancer: the hidden consequence of modern imaging
Katherine Vallecilla Valencia . Herney Andres Garcia‑Perdomo
International Urology and Nephrology https://doi.org/10.1007/s11255-026-05173-6 - Active surveillance (AS) has become a validated management option for small renal masses (SRMs). Prospective data from the DISSRM Registry and pooled analyses demonstrate cancer-specific survival rates exceeding 98%, with metastatic progression observed in fewer than 2% of patients. Suitable candidates typically present with:
• Lesions ≤ 4 cm (T1a)
• Low anatomical complexity (RENAL or PADUA < 7)
• Advanced age or comorbidities limiting surgical benefit.
Overdiagnosis in renal cancer: the hidden consequence of modern imaging
Katherine Vallecilla Valencia . Herney Andres Garcia‑Perdomo
International Urology and Nephrology https://doi.org/10.1007/s11255-026-05173-6 - Recognizing overdiagnosis is not a call to abandon early detection but to refine it—to focus on biologically and clinically meaningful disease. Through active surveillance, Aienhanced diagnostics, molecular profiling, and imaging stewardship, clinicians can strike a balance between detection and discernment. The overarching goal is precision oncology that detects the right cancer in the right patient at the right time.
Overdiagnosis in renal cancer: the hidden consequence of modern imaging
Katherine Vallecilla Valencia . Herney Andres Garcia‑Perdomo
International Urology and Nephrology https://doi.org/10.1007/s11255-026-05173-6
Musculoskeletal
- Objective To explore the associations of CT-evaluated body composition with early recurrence (ER) and overall survival (OS) in patients with pancreatic ductal adenocarcinoma (PDAC) after resection.
Conclusion High VSR was an independent predictor for ER and worse OS in PDAC. Moreover, combining body composition metrics and clinicopathological indicators can improve the prognosis prediction of patients with PDAC after surgery.
Prognostic value of body composition on early recurrence and long-term survival of resectable pancreatic ductal adenocarcinoma
Linxia Wu, Tong Nie, Xiaoling Zhi et al.
Eur Radiol. 2026 Apr;36(4):2945-2964 - Question What are the associations of CT-evaluated body composition with early recurrence and overall survival in patients with pancreatic ductal adenocarcinoma after resection?
Findings High visceral-to-subcutaneous fat ratio is an independent predictor for early recurrence, whereas high skeletal muscle density and subcutaneous fat area independently predict better overall survival.
Prognostic value of body composition on early recurrence and long-term survival of resectable pancreatic ductal adenocarcinoma
Linxia Wu, Tong Nie, Xiaoling Zhi et al.
Eur Radiol. 2026 Apr;36(4):2945-2964 - This study offers important clinical implications for staging, treatment planning, surgical decision-making, and future research. The findings may influence the decision to adopt a “surgery-first” approach or neoadjuvant chemotherapy, particularly in high-risk patients with unfavorable body composition profiles, such as low SMD or high VSR. The reversibility of skeletal muscle deterioration and visceral fat accumulation offers an opportunity for preoperative rehabilitation to enhance patient outcomes and improve survival outcomes. Furthermore, the developed model is useful for prognostic prediction inpatients with pancreatic cancer, aiding in more accurate risk stratification and personalized treatment strategies.
Prognostic value of body composition on early recurrence and long-term survival of resectable pancreatic ductal adenocarcinoma
Linxia Wu, Tong Nie, Xiaoling Zhi et al.
Eur Radiol. 2026 Apr;36(4):2945-2964
OB GYN
- Commonly seen acute complications after cesarean delivery include subfascial and bladder flap hematomas, uterine dehiscence or rupture, endometritis, septic thrombophlebitis, retained products of conception, and vascular complications such as uterine artery pseudoaneurysm, arteriovenous fistula, and ovarian vein thrombosis. On the contrary, delayed complications include uterine scar niche, cesarean scar ectopic pregnancy, endometriosis, pelvic inflammatory disease, bowel obstruction, that occur beyond the post-partum period and are often manifested during subsequent pregnancies.
Early and delayed post-cesarean complications: an imaging review
Nicholas A. Zacharias et al.
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-025-05118-0 - Following cesarean delivery, expected postoperative changes surrounding the hysterotomy site include hypo-enhancement, development of a small hematoma (< 4 cm in size), and/or mild peri-uterine and peri-incisional edema. During the first week, the hysterotomy scar often demonstrates a triangular or oval-shaped isoechoic to hypoechoic area on US and hypodense area on CT along the anterior wall of the uterus, best seen in the sagittal plane . The hysterotomy site then gradually evolves as blood products resorb and a fibrous scar forms with a variably sized uterine wall defect [9]. Small amounts of pelvic fluid and peri-incisional fat stranding and haziness can be seen at the site of hysterotomy and in the anterior abdominal wall .
Early and delayed post-cesarean complications: an imaging review
Nicholas A. Zacharias et al.
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-025-05118-0 - The site of hematomas can be classified using an outside-in anatomic approach (i.e., subcutaneous, rectus sheath, subfascial, bladder flap hematomas with pelvic or retroperitoneal extension) . Often, there are hematomas at more than one site. Damage to the inferior epigastric vessels or its branches cause subfascial and rectus sheath hematomas. Rectus sheath hematoma can be seen as ipsilateral enlargement along the rectus muscle and carries the risk of extension to the pre-vesical space. Subfascial hematoma is characterized by the accumulation of blood in the prevesicle space, posterior to the rectus sheath and anterior to the peritoneum. This type of hematoma carries a higher risk of blood loss due to potential extension into the retroperitoneum.
Early and delayed post-cesarean complications: an imaging review
Nicholas A. Zacharias et al.
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-025-05118-0 - Endometritis is the most common post-partum infection. Cesarean delivery is the greatest risk factor . The diagnosis is primarily clinical with common symptoms including fever, abnormal vaginal discharge, uterine tenderness, and persistent uterine enlargement. Imaging findings are relatively non-specific. US may demonstrate an enlarged uterus with echogenic blood products and echogenic foci with posterior dirty shadowing from gas-producing bacteria. Doppler may show hypervascular endometrium. Similarly, CT may show an enlarged uterus with fluid, air, and debris and endometrial thickening and hyperenhancement . Endometritis can progress to salpingitis and pyometra
Early and delayed post-cesarean complications: an imaging review
Nicholas A. Zacharias et al.
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-025-05118-0 - Symptomatic ovarian vein thrombosis (OVT) is seen in 0.01–0.05% of all deliveries and is more common post-cesarean delivery. Ovarian vein thrombosis can be complicated by septic thrombophlebitis, extension of the thrombus into the left renal vein, the inferior vena cava, and rarely development of pulmonary emboli . A vast majority of these cases are right-sided, due to greater venous stasis and greater predilection for ascending infection. CT is the first line and preferred modality to assess the extent of involvement of the vein. On CT or MRI, OVT with thrombophlebitis is seen as an intraluminal filling defect in the ovarian vein with variable peri-vascular fat stranding and venous wall hyperenhancement. The pelvic veins may also be engorged due to venous stasis.
Early and delayed post-cesarean complications: an imaging review
Nicholas A. Zacharias et al.
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-025-05118-0
Pancreas
Cinematic rendering in pancreatic imaging: CT features and malignancy indicators of mucinous cystic neoplasms.
Rahmatullah ZF, Krueger S, Smith CW, Soyer P, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 20. doi: 10.1007/s00261-026-05568-0..- Pancreatic mucinous cystic neoplasms (MCNs) are benign lesions with potential for malignant transformation. Accurate differentiation from other pancreatic cystic lesions remains challenging through conventional imaging modalities. Cinematic rendering (CR) is an advanced 3D computed tomography (CT) post-processing technique that generates photorealistic images with enhanced texture mapping and improved anatomical detail. It has not been previously discussed as a diagnostic tool for MCNs. Through added depth perception and tissue differentiation, CR complements conventional CT by enhancing visualization of morphology, internal components such as septations, calcification patterns, vascular involvement, and features suspicious of malignancy. It also helps with assessment of resectability and surgical planning by clearly highlighting vascular involvement. It is of note that further research is needed to quantify its added diagnostic value and optimize its role in clinical decisions in the evaluation of MCNs.
- Cinematic rendering in pancreatic imaging: CT features and malignancy indicators of mucinous cystic neoplasms.
Rahmatullah ZF, Krueger S, Smith CW, Soyer P, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 20. doi: 10.1007/s00261-026-05568-0.. - MCNs demonstrate a female predominance, occurring in approximately 84.5% of cases, with a mean age at diagnosis between 40 and 60 years. However, male patients with MCNs exhibit significantly higher malignancy rates (47.2%) compared to females (16.6%). As for location, MCNs are most commonly found in the pancreatic body and tail, accounting for 93–95% of cases. The prevalence of invasive carcinoma arising from MCNs ranges between 6 and 55%, and can be present even in asymptomatic patients . The clinical presentation of MCNs range from an incidental finding to symptoms such as epigastric fullness, a palpable abdominal mass, or abdominal and back pain.
Cinematic rendering in pancreatic imaging: CT features and malignancy indicators of mucinous cystic neoplasms.
Rahmatullah ZF, Krueger S, Smith CW, Soyer P, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 20. doi: 10.1007/s00261-026-05568-0.. - Advanced imaging techniques, including three-dimensional (3D) visualization, may show potential to better characterize pancreatic cystic lesions. Cinematic rendering (CR) is a recently implemented 3D CT post-processing technique that generates photorealistic images with enhanced texture mapping and improved depiction of anatomical detail compared to conventional two-dimensional imaging. To our knowledge, while CR imaging features of other cystic lesions and malignancies have been reported, those specific to MCNs have not been previously described in detail in the literature.
Cinematic rendering in pancreatic imaging: CT features and malignancy indicators of mucinous cystic neoplasms.
Rahmatullah ZF, Krueger S, Smith CW, Soyer P, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 20. doi: 10.1007/s00261-026-05568-0.. - Notably, there is currently a lack of robust statistical data and comprehensive comparative research assessing the clinical utility and accuracy of CR. Large scale studies are needed to validate its potential both as an adjunct to, and in direct comparison with conventional CT. Moreover, achieving the full potential of advanced 3D techniques like CR requires sufficient contrast enhancement, high quality source images, dedicated software, and radiologists who are experts in these methods. Differences in imaging protocols, software iterations, and reconstruction parameters may also limit reproducibility. Additionally, any patient or scanner based artifacts are reflected in the 3D reconstruction.
Cinematic rendering in pancreatic imaging: CT features and malignancy indicators of mucinous cystic neoplasms.
Rahmatullah ZF, Krueger S, Smith CW, Soyer P, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 20. doi: 10.1007/s00261-026-05568-0.. - CR is routinely used to evaluate pancreatic lesions because it provides a comprehensive view of pancreatic morphology and increased tissue contrast and shadowing can potentially help with the characterization of pancreatic cystic neoplasms. The lighting mode of cinematic rendering uses multiple light sources and reviews the images as a volume, not as a single slice, and the user controls the size of the volume display. This volume display is especially valuable when looking at different tissue types and is instrumental when creating vascular maps for accurate tumor staging or preoperative planning. Drawing on our extensive experience, we have developed optimized rendering parameters tailored for pancreatic pathology.
Cinematic rendering in pancreatic imaging: CT features and malignancy indicators of mucinous cystic neoplasms.
Rahmatullah ZF, Krueger S, Smith CW, Soyer P, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 20. doi: 10.1007/s00261-026-05568-0.. - MCNs are macrocystic lesions of the pancreas, that are characterized by a unilocular cyst or a mass composed of a few larger cysts (> 2 cm), ranging from 1.5 to 15 cm in size. MCNs most commonly originate from the body and tail of the pancreas and can exhibit an exophytic growth pattern. They typically demonstrate a round or oval shape with smooth, well-defined borders that often lack external lobulations. A characteristic feature of MCNs is the presence of a thick fibrotic wall that demonstrates enhancement on contrast-enhanced CT images, and a lack of invasion of surrounding structures or vessels.
Cinematic rendering in pancreatic imaging: CT features and malignancy indicators of mucinous cystic neoplasms.
Rahmatullah ZF, Krueger S, Smith CW, Soyer P, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 20. doi: 10.1007/s00261-026-05568-0.. - Importantly, MCNs do not communicate with the main pancreatic duct. This is a key feature that distinguishes them from IPMNs. CR’s inherent ability to display spatial relationships between structures can help depict the separation between the pancreatic duct and cyst. This anatomical detail can help illustrate the separation of duct and cyst that define MCNs, or reveal any ductal involvement seen in lesions such as SCAs or IPMNs. MCNs appear hypodense relative to the surrounding pancreatic parenchyma on CT, with fluid range internal attenuation that reflects their mucinous contents. However, complex fluid density may also be observed due to internal hemorrhage or debris.
Cinematic rendering in pancreatic imaging: CT features and malignancy indicators of mucinous cystic neoplasms.
Rahmatullah ZF, Krueger S, Smith CW, Soyer P, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 20. doi: 10.1007/s00261-026-05568-0.. - The chances of malignancy, including high-grade dysplasia and invasive carcinoma, increase as cyst complexity increases. This complexity may manifest as thickened septations, wall irregularity, larger cyst size, and internal enhancing solid components such as mural nodules. These features may be subtle on conventional imaging but need to be carefully evaluated. CR can potentially improve the detection of malignant changes through its enhanced surface shadows and dynamic textural visualization, making features such as septal thickening, wall irregularity, and internal density changes more visible. These changes in attenuation are rendered as differing visual contrasts on CR, which can help identify internal complexity that might otherwise be overlooked on conventional CT .This can potentially lead to early detection of invasive carcinoma.
Cinematic rendering in pancreatic imaging: CT features and malignancy indicators of mucinous cystic neoplasms.
Rahmatullah ZF, Krueger S, Smith CW, Soyer P, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 20. doi: 10.1007/s00261-026-05568-0.. - The chances of malignancy, including high-grade dysplasia and invasive carcinoma, increase as cyst complexity increases. This complexity may manifest as thickened septations, wall irregularity, larger cyst size, and internal enhancing solid components such as mural nodules. These features may be subtle on conventional imaging but need to be carefully evaluated. CR can potentially improve the detection of malignant changes through its enhanced surface shadows and dynamic textural visualization, making features such as septal thickening, wall irregularity, and internal density changes more visible. These changes in attenuation are rendered as differing visual contrasts on CR, which can help identify internal complexity that might otherwise be overlooked on conventional CT .This can potentially lead to early detection of invasive carcinoma.
Cinematic rendering in pancreatic imaging: CT features and malignancy indicators of mucinous cystic neoplasms.
Rahmatullah ZF, Krueger S, Smith CW, Soyer P, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 20. doi: 10.1007/s00261-026-05568-0.. - MCNs of the pancreas pose a diagnostic challenge as their imaging features overlap with those of other cystic lesions and diagnosis based on imaging can be difficult. However, accurate identification is essential given that they are premalignant lesions. Some characteristic CT features of MCNs include macrocystic architecture, thick enhancing walls, smooth contours, absence of main pancreatic duct communication, and location in the pancreatic body and tail. The 3D visualization that CR provides can help identify these features, supplementing conventional CT in the evaluation of MCNs. Furthermore CR has to potential to help detect features suggestive of malignancy such as enhancing mural nodules, irregular walls, complex internal contents, and calcifications.
Cinematic rendering in pancreatic imaging: CT features and malignancy indicators of mucinous cystic neoplasms.
Rahmatullah ZF, Krueger S, Smith CW, Soyer P, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 20. doi: 10.1007/s00261-026-05568-0..
- Purpose To evaluate the value of a combined model based on multiphase contrast-enhanced CT radiomics and clinical features for differentiating hypovascular pancreatic neuroendocrine tumors (hypo-PNETs) from pancreatic ductal adenocarcinoma (PDAC). Conclusion A combined model integrating multiphasic CT radiomics and clinical features showed promising performance for differentiating hypo-PNETs from PDAC. This model may provide complementary support for preoperative diagnosis, although its incremental value over the radiomics-only model requires further validation.
Radiomics based on multiphasic contrast-enhanced CT combined with clinical features for differentiating hypovascular pancreatic neuroendocrine tumors and pancreatic ductal adenocarcinoma
Fan Xia · Jie Yu · Jianhua Wang · Zhongqiu Wang
J Cancer Res Clin Oncol. 2026 Apr 25;152(4):96. - Pancreatic duct dilatation, tumor composition, age, and maximum tumor diameter were identified as independent predictors. Among the radiomics models, the combined three-phase SVM model achieved the best performance, with AUCs of 0.814 and 0.812 in the training and test sets, respectively. The combined model yielded the highest AUCs (0.886 in the training set and 0.849 in the test set); however, because several between-model comparisons in the test set did not reach statistical significance, its advantage should be interpreted as a potential incremental benefit rather than definitive superiority.
Radiomics based on multiphasic contrast-enhanced CT combined with clinical features for differentiating hypovascular pancreatic neuroendocrine tumors and pancreatic ductal adenocarcinoma
Fan Xia · Jie Yu · Jianhua Wang · Zhongqiu Wang
J Cancer Res Clin Oncol. 2026 Apr 25;152(4):96. - From a translational perspective, the combined model may serve as a decision-support tool rather than a standalone diagnostic system. In future work, it could be converted into a nomogram or a web-based calculator to facilitate clinical application. Because contrast-enhanced CT is routinely available, such a model may be particularly useful in centers with limited expertise in pancreatic tumor imaging, where it could provide complementary information for radiologists and multidisciplinary teams when conventional imaging findings overlap.
Radiomics based on multiphasic contrast-enhanced CT combined with clinical features for differentiating hypovascular pancreatic neuroendocrine tumors and pancreatic ductal adenocarcinoma
Fan Xia · Jie Yu · Jianhua Wang · Zhongqiu Wang
J Cancer Res Clin Oncol. 2026 Apr 25;152(4):96.
- Pancreatic cancer is currently the third leading cause of cancer-related death in the United States, and death rates have gradually increased from 5 per 100,000 in both men and women in the 1930s to 13 and 10 per 100,000 in men and women, respectively, in 2025. In 2024, there were over 66,000 new cases and 51,000 deaths reported in the United States, with global incidence exceeding half a million. Mortality continues to rise, and by 2030, pancreatic cancer is -projected to become the second leading cause of cancer death after lung cancer.
Pancreatic Cancer Screening and Early Detection
Current Strategies and Emerging Innovations in Imaging
Shravya Srinivas Rao , Ok Kyu Song, Yoshifumi Noda et al.
Radiol Clin N Am 64 (2026) 489–499 - A shared consensus among these guidelines is that surveillance should be considered for individuals with an estimated lifetime risk exceeding 5%. 5,7,8 Individuals at elevated risk for pancreatic cancer generally fall into 2 categories: those with a known germline mutation associated with an inherited cancer syndrome, and those with familial pancreatic cancer, defined by a familial aggregation of pancreatic cancer in the absence of an identifiable hereditary cancer syndrome.
Pancreatic Cancer Screening and Early Detection
Current Strategies and Emerging Innovations in Imaging
Shravya Srinivas Rao , Ok Kyu Song, Yoshifumi Noda et al.
Radiol Clin N Am 64 (2026) 489–499 - Mucinous cystic neoplasm (MCN) of pancreas are mucin-producing cystic lesions characterized by the presence of ovarian-type stroma, lacking communication with the pancreatic duct. Management of MCN is guided by lesion size, highrisk imaging features, symptoms, and surgical candidacy. While the AGA and American College of Gastroenterology support surveillance for asymptomatic cysts less than 3 cm without highrisk features (eg, enhancing mural nodules, ductal dilation, or elevated CA 19–9), they advise MR imaging at 1 year and biennial imaging for up to 5 years if stable. Surveillance is not recommended for patients unfit for surgery. Surgical resection remains the standard for symptomatic or high-risk lesions despite its size.
Pancreatic Cancer Screening and Early Detection
Current Strategies and Emerging Innovations in Imaging
Shravya Srinivas Rao , Ok Kyu Song, Yoshifumi Noda et al.
Radiol Clin N Am 64 (2026) 489–499 - The incidence of pancreatic cancer in branch-duct IPMN (BD-IPMN) is 2% to 10% and in main duct IPMN is much higher, with rates varying from 6% to 45%. 26,27 This increased risk prompts guidelines to recommend that continued surveillance to be considered for patients with small, unchanged IPMNs to identify the early transformation of benign cysts to highgrade dysplasia or invasive carcinoma. 24 These include features of high-risk stigmata and worrisome features. High risk stigmata includes obstructive jaundice, enhancing mural nodule greater than 5 mm, and main pancreatic duct (MPD) dilatation greater than 10 mm.
Pancreatic Cancer Screening and Early Detection
Current Strategies and Emerging Innovations in Imaging
Shravya Srinivas Rao , Ok Kyu Song, Yoshifumi Noda et al.
Radiol Clin N Am 64 (2026) 489–499
- Renal cell carcinoma (RCC) is the most common primary tumor causing both solitary and multiple pancreatic metastasis, which can occur years after initial diagnosis. The mean time interval from RCC to pancreatic metastases is greater than 10 years, and a period as long as 32.7 years has been described.
The imaging dilemma of solid serous cystadenomas, PNETs, and vascular pancreatic metastases: Challenges and diagnostic strategies.
Rahmatullah ZF, Arshad H, Fishman EK.
Curr Probl Diagn Radiol. 2026 Jun 8:S0363-0188(26)00111-8. 
The imaging dilemma of solid serous cystadenomas, PNETs, and vascular pancreatic metastases: Challenges and diagnostic strategies.
Rahmatullah ZF, Arshad H, Fishman EK.
Curr Probl Diagn Radiol. 2026 Jun 8:S0363-0188(26)00111-8.- Solid SCAs are a rare variant of serous cystadenomas that pose a diagnostic challenge on cross-sectional imaging due to their atypical appearance. Unlike typical cystic appearance of SCAs, solid SCAs exhibit imaging features that often overlap with other enhancing pancreatic masses, such as PNETs and metastatic lesions. Although patient demographics, clinical history of prior malignancy and enhancement characteristics may aid in distinguishing these entities, the accurate differentiation between solid SCAs, PNETs and metastatic lesions solely based on imaging remains difficult. Biopsy with histopathology is required for definitive diagnosis. The distinction is crucial, as SCAs are benign, whereas PNETs and metastases are malignant and often necessitate more invasive management strategies. Therefore, solid SCA should be considered in the differential diagnosis of a hypervascular solid pancreatic mass, particularly when no primary malignancy is known, to avoid unnecessary interventions, although radiologists must remain mindful of the substantial imaging overlap.
The imaging dilemma of solid serous cystadenomas, PNETs, and vascular pancreatic metastases: Challenges and diagnostic strategies.
Rahmatullah ZF, Arshad H, Fishman EK.
Curr Probl Diagn Radiol. 2026 Jun 8:S0363-0188(26)00111-8.
- Solid serous cystadenomas (SCAs) of the pancreas represent a rare subtype that can closely mimic malignant pancreatic lesions on imaging, leading to potential diagnostic pitfalls. While SCAs are typically benign cystic masses, the solid variant exhibits imaging patterns similar to some pancreatic neuroendocrine tumors (PNETs) and metastatic lesions, complicating their differentiation on cross-sectional modalities such as CT. In this pictorial review, we analyze both standard two-dimensional (2D) CT images and advanced three-dimensional (3D) cinematic rendering techniques to compare imaging features of solid SCAs with those of PNETs and pancreatic metastases. Accurate recognition is essential to avoid unnecessary surgical intervention, particularly since SCAs do not require resection, unlike their malignant counterparts.
The imaging dilemma of solid serous cystadenomas, PNETs, and vascular pancreatic metastases: Challenges and diagnostic strategies.
Rahmatullah ZF, Arshad H, Fishman EK.
Curr Probl Diagn Radiol. 2026 Jun 8:S0363-0188(26)00111-8. - Pancreatic serous cystadenomas (SCAs) are classically cystic lesions and represent 10–16% of cystic pancreatic lesions.1 Clinical characteristics, analysis of cyst fluid, endoscopic ultrasound (EUS), and advanced cross-sectional imaging modalities are valuable methods for distinguishing among pancreatic cystic lesions. It is particularly important to accurately recognize SCAs, as they are benign masses that often do not necessitate surgical intervention and may exhibit distinctive features that aid in their diagnosis. SCAs are frequently discovered incidentally during imaging studies, as nearly 60% of patients are asymptomatic. When symptoms do occur, they tend to be non-specific and may include abdominal pain, abdominal mass, or other pancreaticobiliary symptoms.They are most commonly found in women aged 60-80 with CT imaging being the modality of choice for evaluation.
The imaging dilemma of solid serous cystadenomas, PNETs, and vascular pancreatic metastases: Challenges and diagnostic strategies.
Rahmatullah ZF, Arshad H, Fishman EK.
Curr Probl Diagn Radiol. 2026 Jun 8:S0363-0188(26)00111-8. - SCAs are categorized based on their morphological patterns: microcystic (numerous small cysts with central scar), macrocystic or oligocystic (cysts > 2 cm), honeycomb (indistinguishable tiny cysts), and rarely, solid. CT imaging helps identify any septations, calcifications, or vascularity, all of which are features typically associated with SCAs. More atypical manifestations include giant SCAs (>10 cm), ductal dilation, intratumoral hemorrhage, and the ability to invade adjacent structures, such as the muscle, vessels, nerves, and lymph nodes..
The imaging dilemma of solid serous cystadenomas, PNETs, and vascular pancreatic metastases: Challenges and diagnostic strategies.
Rahmatullah ZF, Arshad H, Fishman EK.
Curr Probl Diagn Radiol. 2026 Jun 8:S0363-0188(26)00111-8. - While typically considered as a benign cystic mass, around 3% of SCAs appear as an enhancing mass and are classified as a solid SCA . Histologically, these lesions lack cystic spaces, and the cells are organized into nests, sheets, and trabeculae that are separated by thick fibrous bands containing numerous capillaries. The stroma exhibits significant contrast enhancement, contributing to the solid hypervascular appearance seen on CT images. These lesions are notoriously difficult to diagnose on imaging because their features overlap with those of other pancreatic lesions. For example, solid SCAs do not present with the cystic structures, septations, or central scar that are typical of SCAs. Rather, they are completely solid on imaging. 3D cinematic rendering can potentially help distinguish solid SCAs from cystic by highlighting texture differences.”
The imaging dilemma of solid serous cystadenomas, PNETs, and vascular pancreatic metastases: Challenges and diagnostic strategies.
Rahmatullah ZF, Arshad H, Fishman EK.
Curr Probl Diagn Radiol. 2026 Jun 8:S0363-0188(26)00111-8. - PNETs are rare malignancies that originate from the neuroendocrine cells in the pancreas, accounting for approximately 3% of all pancreatic tumors.These tumors can occur in individuals of any age, although the median age is in the sixth decade of life, with a male-to-female ratio of approximately 1:1. On CT imaging, these tumors appear as hypervascular pancreatic masses with either a solid, mixed cystic or even completely cystic appearance, potentially exhibiting internal necrosis or hemorrhage. PNETs can be classified into two groups based on hormone production: functioning (F-PNET) and non-functioning (NF-PNET). Functioning PNETs present with hormone-related clinical symptoms, a feature specific to PNETs, as endocrine disturbances such as hypoglycemia were reported exclusively in the PNET group compared to SCAs in a comparative study.
The imaging dilemma of solid serous cystadenomas, PNETs, and vascular pancreatic metastases: Challenges and diagnostic strategies.
Rahmatullah ZF, Arshad H, Fishman EK.
Curr Probl Diagn Radiol. 2026 Jun 8:S0363-0188(26)00111-8. - Secondary malignancies to the pancreas can arise through hematogenous spread (75%), systemic disease like lymphoma (20%), or from direct invasion by a primary malignancy arising from an adjacent organ (5%). Patients with pancreatic metastases are typically detected during the initial evaluation of the primary tumor, on routine surveillance following resection of the primary tumor, or when new symptoms arise due to the pancreatic lesion.A study of 127 patients with metastatic cancer to the pancreas reported that the majority of the patients were asymptomatic (27.65%), with common symptomatic presentations including jaundice (25.2%) and abdominal pain (19.7%).The most common site of primary tumor causing metastasis to the pancreas is the kidney (70%), followed by breast, lung, colorectal, and skin.Imaging appearances of the metastases varies based on the primary tumor. They may present as solitary, multifocal, or diffuse deposits, mostly appearing as well-defined, discrete lesions with smooth margins. They are often hyperenhancing on imaging, raising concern for solid SCAs and PNETs, but patient history of prior primary cancer is the most important clue to the diagnosis of metastases.
The imaging dilemma of solid serous cystadenomas, PNETs, and vascular pancreatic metastases: Challenges and diagnostic strategies.
Rahmatullah ZF, Arshad H, Fishman EK.
Curr Probl Diagn Radiol. 2026 Jun 8:S0363-0188(26)00111-8. - The main differences concern how early EUS enters the algorithm and how willing each guideline is to stop follow-up. The AGA is the most conservative, relying predominantly on MRI and reserving EUS-FNA for cysts with at least two high-risk features or interval change. The ACG also uses cross-sectional imaging as the surveillance backbone but adopts a more flexible approach and brings EUS earlier into the pathway when symptoms, CA 19-9 elevation, mural nodularity, ductal changes, or rapid growth emerge . The European and IAP frameworks are the most explicitly risk stratified. Fukuoka 2017 uses MRI or CT as the basis of follow-up and escalates to EUS when worrisome features are present, whereas Kyoto 2024 retains this overall structure but places EUS, contrast-enhanced EUS, and positive cytology obtained by EUS more centrally within the formal risk-classification scheme.
Evidence-based approach to the diagnosis, management and surveillance of pancreatic cystic lesions: from the guidelines to the clinical practice.
Gincul R, Napoleon B.
Best Pract Res Clin Gastroenterol. 2026 Mar; in press. - Postoperative surveillance reveals the clearest histology-based divergence. The AGA again takes the least intensive position, recommending MRI every 2 years only after resection showing invasive cancer or dysplasia, and no routine surveillance when resection shows neither high-grade dysplasia nor malignancy. In contrast, the ACG, Fukuoka, European, and Kyoto guidelines all support continued surveillance after IPMN resection because recurrence, multifocal disease, and metachronous PDAC remain relevant even after non-invasive disease. Follow-up is intensified in the presence of high-grade dysplasia, main-duct IPMN, positive or high-risk margins, family history of PDAC, or residual pancreatic lesions. By contrast, all guidelines agree that completely resected MCNs without associated adenocarcinoma do not require further surveillance .
Evidence-based approach to the diagnosis, management and surveillance of pancreatic cystic lesions: from the guidelines to the clinical practice.
Gincul R, Napoleon B.
Best Pract Res Clin Gastroenterol. 2026 Mar; in press. - IMPORTANCE Low-risk pancreatic cystic lesions (PCLs) represent common incidental findings with potential for malignant transformation, warranting long-term surveillance. However, data on their long-term cancer risk are limited, leading to inconsistency in current surveillance strategies.
OBJECTIVE To determine the long-term incidence of pancreatic cancer among patients with low-risk PCLs, and to identify baseline clinical and imaging factors associated with cancer development.
Pancreatic Cancer Risk in Patients With Low-Risk Cystic Lesions.
Haj Mirzaian et al.
JAMA Netw Open. 2026 May 1;9(5):e2613808. - Among 499,631 patients reviewed, 6064 with low-risk PCLs were identified and included in the analytic sample, contributing 20 145 person-years of follow-up. Patients had a mean (SD) age at diagnosis of 65.9 (12.3) years and included 3612 females (59.6%). Of the 6064 patients, 38 (0.6%) developed pancreatic cancer, with an incidence rate of 1.89 (95%CI, 1.29-2.49) cases per 1000 person-years, which was higher than the previously reported general population rate of 0.14 cases per 1000 person-years. Twenty-six of 38 patients (68.4%) had cancer that arose from the cyst site, while 12 (31.6%) developed cancer from a different region of the pancreas. Ten patients (26.3%) were diagnosed more than 5 years after initial PCL detection. In multivariable analysis, larger cyst size (HR, 2.24; 95%CI, 1.45-3.48), main pancreatic duct ectasia (HR, 2.84; 95%CI, 1.18-6.84), and older age (HR, 1.04; 95%CI, 1.01-1.07) were associated with cancer. Adding age to a cyst size–based risk stratification model improved estimation of pancreatic cancer risk (NRI, 0.20 [95%CI, 0.03-0.37]; change in C-statistic, 0.14 [95%CI, 0.07-0.22]).
Pancreatic Cancer Risk in Patients With Low-Risk Cystic Lesions.
Haj Mirzaian et al.
JAMA Netw Open. 2026 May 1;9(5):e2613808. - CONCLUSIONS AND RELEVANCE This cohort study found that low-risk PCLs were associated with a sustained long-term pancreatic cancer risk and were best stratified by combining clinical andimaging factors. Longer than a 5-year follow-up of low-risk PCLs may be warranted to reduce missed or delayed diagnosis of pancreatic cancer.
Pancreatic Cancer Risk in Patients With Low-Risk Cystic Lesions.
Haj Mirzaian et al.
JAMA Netw Open. 2026 May 1;9(5):e2613808. - In this retrospective multisite cohort study of patients with low-risk PCLs, larger cyst size, MPD ectasia, and older age were independently associated with pancreatic cancer development. Incorporating age into cyst size–based surveillance strategies was associated with improved risk stratification. Given that 26.3%of pancreatic cancers were diagnosed beyond 5 years of follow-up, longer than a 5-year follow-up of low-risk PCLs may be needed to reduce missed or delayed diagnosis.
Pancreatic Cancer Risk in Patients With Low-Risk Cystic Lesions.
Haj Mirzaian et al.
JAMA Netw Open. 2026 May 1;9(5):e2613808.
- IMPORTANCE Artificial intelligence (AI) is changing health and health care on an unprecedented scale. Though the potential benefits are massive, so are the risks. The JAMA Summit on AI discussed how health and health care AI should be developed, evaluated, regulated, disseminated, and monitored.
CONCLUSIONS AND RELEVANCE AI will disrupt every part of health and health care delivery in the coming years. Given the many long-standing problems in health care, this disruption represents an incredible opportunity. However, the odds that this disruption will improve health for all will depend heavily on the creation of an ecosystem capable of rapid, efficient, robust, and generalizable knowledge about the consequences of these tools on health.
AI, Health, and Health Care Today and Tomorrow
The JAMA Summit Report on Artificial Intelligence
Derek C. Angus et al.
JAMA. 2025;334(18):1650-1664. - Ensuring AI is deployed equitably and in a manner that improves health outcomes or, if improving efficiency of health care delivery, does so safely, requires progress in 4 areas. First, multistakeholder engagement throughout the total product life cycle is needed. This effort would include greater partnership of end users with developers in initial tool creation and greater partnership of developers, regulators, and health care systems in the evaluation of tools as they are deployed. Second, measurement tools for evaluation and monitoring should be developed and disseminated. Beyond proposed monitoring and certification initiatives, this will require new methods and expertise to allow health care systems to conduct or participate in rapid, efficient, and robust evaluations of effectiveness. The third priority is creation of a nationally representative data infrastructure and learning environment to support the generation of generalizable knowledge about health effects of AI tools across different settings. Fourth, an incentive structure should be promoted, using market forces and policy levers, to drive these changes.
AI, Health, and Health Care Today and Tomorrow
The JAMA Summit Report on Artificial Intelligence
Derek C. Angus et al.
JAMA. 2025;334(18):1650-1664. - The following 3 advances help differentiate AI from prior digital technologies:
• Deep learning: development of deeper, more convoluted neural networks capable of interpreting large complex datasets to address specific yet complicated tasks (eg, computer vision).
• Generative AI: an extension of deep learning using so-called large language and foundation models capable of generating new content to address far broader task requests (eg, ChatGPT or Gemini).
• Agentic AI: an extension of deep learning and generative A capable of autonomous decision-making (eg, the Tesla autopilotsoftware for autonomous driving).
AI, Health, and Health Care Today and Tomorrow
The JAMA Summit Report on Artificial Intelligence
Derek C. Angus et al.
JAMA. 2025;334(18):1650-1664. - Health and health care AI tools should be subject to a governance structure that protects individuals and ensures the tools achieve their potential benefits. For other health care interventions, regulatory oversight is an important part of that governance, assuring society and markets that an intervention is credible. However, the US has no comprehensive fit-for-purpose regulatory framework for health and health care AI. Reasons include the diverse and rapidly evolving nature of AI technology, the numerous agencies with jurisdiction over different types and aspects of AI, and a lack of regulatory frameworks specifically tailored forAI within these agencies.41 Drug and traditional medical device development also benefits from international harmonization of regulatory standards.
- Finally, use of AI tools has thus far largely been voluntary. However, as their benefits become more established, failure to useanAItoolmaybeconsideredunethicalorabreachofstandardcare.Ahealth care system or professional may thus be liable in a malpractice suit for failing to use AI. At the same time, if a plaintiff sues for an adverse outcome when care was provided in which an AI tool was involved, the question arises of whether liability rests with the health care professional, the health care system,or the developer of the tool. Though relevant case law is currently limited, developers, health care systems, and health care professionals will all have to adopt strategies to manage their liability risk. These examples are just some of the many new issues that will need to be addressed as Aibecomes more incorporated in health and health care.
AI, Health, and Health Care Today and Tomorrow
The JAMA Summit Report on Artificial Intelligence
Derek C. Angus et al.
JAMA. 2025;334(18):1650-1664. - AI will massively disrupt health and health care delivery in the coming years. The traditional approaches to evaluate, regulate,and monitor novel health care interventions are being pushed to their limits, especially with generative and agentic AI, and especially because the tools’ effects cannot be fully understood until deployed in practice. Nonetheless,many tools are already being rapidly adopted,in part because they are addressing important pain points for end users. Given the many long-standing problems in health care, this disruption represents an incredible opportunity. However, the odds that this disruption will improve health for all will depend heavily on creation of an ecosystem capable of rapid, efficient, robust, and generalizable knowledge about the consequences of these tools on health.
AI, Health, and Health Care Today and Tomorrow
The JAMA Summit Report on Artificial Intelligence
Derek C. Angus et al.
JAMA. 2025;334(18):1650-1664. - Pancreatic cystic lesions (PCLs) are increasingly detected due to the widespread use of cross-sectional imaging and represent a significant diagnostic challenge because of their heterogeneous biological behavior, ranging from benign lesions to neoplasms with malignant potential. Accurate characterization and risk stratification are essential to guide appropriate management and avoid unnecessary surgical interventions. Conventional imaging modalities, including computed tomography (CT), magnetic resonance (MR) imaging, and endoscopic ultrasound (EUS), remain central to the diagnostic work-up; however, their ability to reliably differentiate cyst subtypes and predict malignant transformation remains limited. In recent years, artificial intelligence (AI) and radiomics have emerged as promising approaches for improving the non-invasive characterization of PCLs by extracting quantitative imaging features beyond those appreciable through visual assessment.
Radiomics and artificial intelligence in pancreatic cyst characterization: future or fiction?
Cesare Maino · Paolo Niccolò Franco · Federica Omboni
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-026-05515-z - This narrative review summarizes the current evidence regarding CT- and MR-based radiomics and AI in pancreatic cyst characterization, focusing on their role in differentiating mucinous from non-mucinous cysts, identifying high-risk intraductal papillary mucinous neoplasms (IPMNs), and supporting clinical decision-making. The potential advantages of these techniques are discussed alongside main methodological limitations, including variability in imaging acquisition protocols, segmentation reproducibility, small and often retrospective datasets, limited external validation, and interpretability of AI-based models. Further multicenter studies, standardized radiomic pipelines, and prospective validation are required before these tools can be reliably integrated into routine clinical practice.
Radiomics and artificial intelligence in pancreatic cyst characterization: future or fiction?
Cesare Maino · Paolo Niccolò Franco · Federica Omboni
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-026-05515-z - CT is often the first modality in which PCLs are incidentally detected, with reported accuracy of approximately 60–75% in distinguishing cysts with or without malignant potential (highest reported sensitivity and specificity of 72% and 74%, respectively). While CT offers high spatial resolution for size measurement of medium-large cysts, it underperforms for small lesion characterization . Advances such as spectral and photon-counting CT may partly address this limitation; however, CT remains inferior to MR and endoscopic ultrasound (EUS) in detecting septa and cyst with main pancreatic duct (MPD) communication.
Radiomics and artificial intelligence in pancreatic cyst characterization: future or fiction?
Cesare Maino · Paolo Niccolò Franco · Federica Omboni
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-026-05515-z - Accurate and reproducible segmentation of the region of interest (ROI) is fundamental to reliable radiomic feature extraction. In PCLs, segmentation is particularly challenging due to irregular lesion morphology, partial volume effects, ill-defined cyst walls, and the presence of septa or mural nodules . Manual segmentation by expert radiologists remains the reference standard but is time-consuming and subject to significant inter- and intra-observer variability. Semi-automated and DL-based segmentation approaches can improve efficiency and reproducibility, though their performance depends heavily on image quality, lesion heterogeneity, and annotated training data availability.
Radiomics and artificial intelligence in pancreatic cyst characterization: future or fiction?
Cesare Maino · Paolo Niccolò Franco · Federica Omboni
Abdominal Radiology 2026 https://doi.org/10.1007/s00261-026-05515-z - Worrisome features include cyst size greater than 3 cm, enhancing mural nodule less than 5 mm, thickened or enhancing walls, MPD dilatation 5 to 9 mm, abrupt change in caliber of pancreatic duct with distal atrophy, lymphadenopathy, increase in serum CA 19-9, cyst growth rate greater than mm/2 years.
Pancreatic Cancer Screening and Early Detection
Current Strategies and Emerging Innovations in Imaging
Shravya Srinivas Rao , Ok Kyu Song, Yoshifumi Noda et al.
Radiol Clin N Am 64 (2026) 489–499 - A multicenter AI tool on CT scans (PANDA model) reached a sensitivity of 93% and specificity of 99% highlighting its potential in large-scale PDAC screening. Radiomics, which extracts quantitative imaging features invisible to human readers, further boosts early detection. Radiomics-based machine learning classifiers can distinguish early cancer from normal tissue with high accuracy. 35 These models can capture subtle textural changes years before the tumor is evident, enabling identification of high-risk individuals even outside known familial risk groups.
Pancreatic Cancer Screening and Early Detection
Current Strategies and Emerging Innovations in Imaging
Shravya Srinivas Rao , Ok Kyu Song, Yoshifumi Noda et al.
Radiol Clin N Am 64 (2026) 489–499 - • Individuals with strong family history or pathogenic germline mutations (eg, breast cancer gene 2 [BRCA2], CDKN2A, serine/threonine kinase 11 [STK11], and serine protease 1 [PRSS1]) have greater than 5% lifetime PDAC risk and should undergo pancreatic cancer screening by annual MRI/MRCP ± EUS surveillance at specialized centers.
• Pancreatic cystic lesions such as MCNs and IPMNs are precursors of pancreatic cancer, necessitating tailored risk-based surveillance strategies.
Pancreatic Cancer Screening and Early Detection
Current Strategies and Emerging Innovations in Imaging
Shravya Srinivas Rao , Ok Kyu Song, Yoshifumi Noda et al.
Radiol Clin N Am 64 (2026) 489–499 - “The pancreatic cancer diagnosis: radiologists meet AI (PANORAMA) study showed that AI models outperformed 68 radiologists (AUC = 0.92 vs. 0.88) across multi-institutional datasets, underscoring robust performance across different settings. Another study by Chen et al. validated a detection model in a nationwide cohort (n = 1,473), reporting an AUC of 0.95 and sensitivity of 89.7%, with preserved performance for tumors <2 cm in diameter. Together, these findings suggest the potential use of AI in detecting pancreatic tumors early.”
From Bench to Bedside: The Path Toward Real-World Translation for Artificial Intelligence in Pancreatic Cancer Detection.
Syailendra EA, Arshad H, Lopez-Ramirez F, Tixier F, Kawamoto S, Fishman EK, Chu LC.
Korean J Radiol. 2026 Jun;27(6):543-554. doi: 10.3348/kjr.2026.0003. PMID: 42225574; PMCID: PMC13236442. - Retrospective studies showed that progressive pre-diagnostic pancreatic changes such as parenchymal atrophy, increased tissue heterogeneity, and ductal dilation preceded the diagnosis of PDAC . Several groups have evaluated whether AI can retrospectively detect PDAC on pre-diagnostic CT (i.e., obtained before the clinical diagnosis of PDAC). Mukherjee et al. analyzed scans obtained up to 3 years before diagnosis (median lead time, 398 days) and developed radiomics-based classifiers that achieved an AUC of 0.98, demonstrating measurable pancreatic texture and shape differences well before visible tumors formed. Using a CNN model, Korfiatis et al. demonstrated promising results (AUC = 0.97) for pre-diagnostic detection, with a median lead time of 475 days.
From Bench to Bedside: The Path Toward Real-World Translation for Artificial Intelligence in Pancreatic Cancer Detection.
Syailendra EA, Arshad H, Lopez-Ramirez F, Tixier F, Kawamoto S, Fishman EK, Chu LC.
Korean J Radiol. 2026 Jun;27(6):543-554. doi: 10.3348/kjr.2026.0003. PMID: 42225574; PMCID: PMC13236442. - Clinician acceptance remains vital for effective use of AI models. Limited interpretability and the black-box nature of the models in generating predictions reduce clinicians’ willingness to rely on AI. Although explainability tools may enhance transparency, their impact on diagnostic performance is variable. Early involvement of radiologists in task definition, error tolerance, and interface design has been shown to improve clinical relevance and usability, supporting higher adoption rates [82,83]. Patient trust is also a critical factor. Surveys show that confidence in AI for diagnosis or treatment is modest, with many patients expressing discomfort and low trust in health systems to use AI responsibly [84,85]. Addressing these issues requires co-design with end users, task explainability, and communication strategies that clearly articulate the role of AI as an assistive and not autonomous component of the diagnostic process.
From Bench to Bedside: The Path Toward Real-World Translation for Artificial Intelligence in Pancreatic Cancer Detection.
Syailendra EA, Arshad H, Lopez-Ramirez F, Tixier F, Kawamoto S, Fishman EK, Chu LC.
Korean J Radiol. 2026 Jun;27(6):543-554. doi: 10.3348/kjr.2026.0003. PMID: 42225574; PMCID: PMC13236442. - Early detection using AI models provides an opportunity to improve the outcomes of pancreatic cancer patients. Multiple studies showed promising results; however, these remain limited to research settings. Progress will require efforts across different aspects, as adoption depends on factors beyond model specifics, including workflow integration, governance, financial commitment, reimbursement, and ultimately, provider and patient trust. Equally important is post-deployment surveillance to monitor model behaviors clinically. The path forward will depend on AI models that not only demonstrate exceptional performance but also operate reliably within the complex and evolving clinical and societal environments that shape pancreatic cancer care.
From Bench to Bedside: The Path Toward Real-World Translation for Artificial Intelligence in Pancreatic Cancer Detection.
Syailendra EA, Arshad H, Lopez-Ramirez F, Tixier F, Kawamoto S, Fishman EK, Chu LC.
Korean J Radiol. 2026 Jun;27(6):543-554. doi: 10.3348/kjr.2026.0003. PMID: 42225574; PMCID: PMC13236442. - Purpose: The purpose of this study was to evaluate the contribution of radiomics features extracted from various pancreatic structures on computed tomography (CT) images, including the main pancreatic duct and cystic lesion, for predicting the pathological grade of intraductal papillary mucinous neoplasms (IPMNs) using machine learning models Conclusion: Integrating CT-based radiomics features from pancreatic ducts and cysts improves classification performance, with main pancreatic duct features being the most contributive predictor of IPMN grade.
Evaluating the role of the main pancreatic duct in intraductal papillary mucinous neoplasm grading: A multi-structure radiomics-based machine learning approach.
Syailendra EA, Lopez-Ramirez F, Tixier F, Blanco A, Arshad H, Yasrab M, Rahmatullah ZF, Javed AA, He J, Lennon AM, Hruban RH, Afghani E, Kawamoto S, Chu LC, Fishman EK.
Diagn Interv Imaging. 2026 Feb (in press) - The purpose of this study was to evaluate the contribution of radiomics features extracted from different pancreatic structures, including the main pancreatic duct (MPD) and cystic lesion, for predicting the pathological grade of IPMNs using machine learning models.
Evaluating the role of the main pancreatic duct in intraductal papillary mucinous neoplasm grading: A multi-structure radiomics-based machine learning approach.
Syailendra EA, Lopez-Ramirez F, Tixier F, Blanco A, Arshad H, Yasrab M, Rahmatullah ZF, Javed AA, He J, Lennon AM, Hruban RH, Afghani E, Kawamoto S, Chu LC, Fishman EK.
Diagn Interv Imaging. 2026 Feb (in press) - Although not significant, the cyst radiomics model outperformed the MPD radiomics model. The improved performance after incorporating MPD features aligns with the groundwork of pancreatic ductal changes in IPMN pathogenesis. High-grade IPMNs are associated with more extensive MPD involvement, inflammatory changes, and architectural distortion that may be captured through quantitative texture analysis.
Evaluating the role of the main pancreatic duct in intraductal papillary mucinous neoplasm grading: A multi-structure radiomics-based machine learning approach.
Syailendra EA, Lopez-Ramirez F, Tixier F, Blanco A, Arshad H, Yasrab M, Rahmatullah ZF, Javed AA, He J, Lennon AM, Hruban RH, Afghani E, Kawamoto S, Chu LC, Fishman EK.
Diagn Interv Imaging. 2026 Feb (in press) - Previous studies have explored the use of radiomics for predicting IPMN grades with variable performance. However, most prior investigations focused on either cyst-derived or pancreas-derived features [20,21,29,32]. We systematically evaluated the individual and combined contributions of radiomics features derived from both the IPMNs lesion and MPD, allowing for a comprehensive understanding of structure-specific information. From the failure analysis, FPs were driven by MPD-derived features, with SHAP analysis showing greater MPD contribution compared to cyst-based texture features that characterized TNs. We found that FPs clustered closer to TPs and statistical analysis showing no difference between FPs and TPs in some radiomics features. This may suggest that errors arise from radiomics similarities between benign lesions and true high-grade lesions, highlighting the limits of imaging-only approaches and the potential role of multimodal integration to improve specificity.
Evaluating the role of the main pancreatic duct in intraductal papillary mucinous neoplasm grading: A multi-structure radiomics-based machine learning approach.
Syailendra EA, Lopez-Ramirez F, Tixier F, Blanco A, Arshad H, Yasrab M, Rahmatullah ZF, Javed AA, He J, Lennon AM, Hruban RH, Afghani E, Kawamoto S, Chu LC, Fishman EK.
Diagn Interv Imaging. 2026 Feb (in press) - Several research directions should emerge from this work. First, development of automated segmentation algorithms for pancreatic structures would enable broader clinical implementation by eliminating the need for manual segmentation. Towards this goal, deep learning approaches show promise for accurate pancreatic anatomy delineation. Second, ensemble modeling approaches combining multiple structure- specific models based on segmentation availability could maximize clinical utility. For instances where MPD segmentation is challenging, cyst-based models could provide fallback predictions. Finally, multi- institutional collaborations would enable larger dataset and external validation to address generalizability concerns, such as the inclusion of other types of pancreatic cyst.
Evaluating the role of the main pancreatic duct in intraductal papillary mucinous neoplasm grading: A multi-structure radiomics-based machine learning approach.
Syailendra EA, Lopez-Ramirez F, Tixier F, Blanco A, Arshad H, Yasrab M, Rahmatullah ZF, Javed AA, He J, Lennon AM, Hruban RH, Afghani E, Kawamoto S, Chu LC, Fishman EK.
Diagn Interv Imaging. 2026 Feb (in press) - In conclusion, this study demonstrated that multi-structure radiomics improved the performance of pre-surgical prediction of IPMN grade compared to morphological feature such as diameter. MPD radiomics provides valuable discriminative information, complementing cyst-based features and contributing to model performance. Further validation and prospective clinical studies are needed to establish the real-world impact.
Evaluating the role of the main pancreatic duct in intraductal papillary mucinous neoplasm grading: A multi-structure radiomics-based machine learning approach.
Syailendra EA, Lopez-Ramirez F, Tixier F, Blanco A, Arshad H, Yasrab M, Rahmatullah ZF, Javed AA, He J, Lennon AM, Hruban RH, Afghani E, Kawamoto S, Chu LC, Fishman EK.
Diagn Interv Imaging. 2026 Feb (in press)
- BACKGROUND: Current therapies offer limited benefit for patients with previously treated metastatic pancreatic ductal adenocarcinoma (mPDAC). Aberrant activation of the RAS pathway is the key driver of PDAC, with oncogenic RAS mutations present in more than 90% of cases. Daraxonrasib is an oral RAS(ON) multiselective, tri-complex inhibitor of the active guanosine triphosphate–bound state of mutant and wild-type RAS.
METHODS: In this phase 3, international, open-label, randomized trial, we randomly assigned patients with previously treated mPDAC to receive daraxonrasib or chemotherapy of the investigator’s choice. The dual primary end points were overall survival andprogression-free survival in the subpopulation of patients with RAS G12 mutations (the RAS G12 population). Key secondary end points included overall survival and progression-free survival in the overall population (which included patients with RAS G12, G13, or Q61 mutations or with no RAS mutation identified) and objective response and patient-reported quality of life in the RAS G12 and overall populations. Safety was also assessed. - RESULTS: A total of 500 patients, including 91.8% with RAS G12 mutations, were randomly assigned to receive daraxonrasib (248 patients) or chemotherapy (252 patients). Themedian overall survival in the RAS G12 population was 13.2 months with daraxonrasib and 6.6 months with chemotherapy, and the median overall survival in the overall population was 13.2 months and 6.7 months, respectively; the hazard ratio was 0.40 in both populations (P<0.001). The median progression-free survival in the RAS G12 population was 7.3 months with daraxonrasib and 3.5 months with chemotherapy, and that in the overall population was 7.2 months and 3.6 months, respectively; the hazard ratios were 0.45 and 0.49, respectively (P<0.001 for both comparisons). Adverse events that occurred after the start of treatment were reported in all the patients in the daraxonrasib group and in 97.7% of those in the chemotherapy group; the incidence of adverse events of grade 3 or higher was 61.8% and 69.6%, respectively. Treatmentrelated adverse events that led to treatment discontinuation occurred in 1.2% of the patients in the daraxonrasib group and in 11.2% of those in the chemotherapy group.
CONCLUSIONS: Among patients with previously treated mPDAC, treatment with daraxonrasib ledto significantly longer overall survival and progression-free survival than chemotherapy - In both the RAS G12 population and the overall population, treatment with daraxonrasib reduced the risk of death by 60%, with a median overall survival of 13.2 months among patients with previously treated mPDAC. Although crosstrial comparisons should be interpreted with caution, previous randomized mPDAC trials of firstline treatment with FOLFIRINOX, gemcitabine plus nab-paclitaxel, and NALIRIFOX (liposomal irinotecan, fluorouracil, leucovorin, and oxaliplatin) have shown median overall survival ranging from 8.5 to 11.1 months.
Daraxonrasib or Chemotherapy in Previously Treated Metastatic Pancreatic Cancer.
O'Reilly EM, Wainberg ZA, Wolpin BM et al.;
N Engl J Med. 2026 May 31. doi: 10.1056/NEJMoa2605555. Epub ahead of print. PMID: 42223072. - In the RASolute 302 trial, once-daily treatment with oral daraxonrasib resulted in significantly longer overall survival and progression-free survival than standard cytotoxic chemotherapy among patients with previously treated mPDAC, more than 90% of whom had RAS G12 mutations. Overall survival results were largely consistent across patient subgroups, and patient-reported end points favored daraxonrasib. The results of this trial support daraxonrasib as a clinically meaningful advance in the treatment of patients with previously treated mPDAC.
Daraxonrasib or Chemotherapy in Previously Treated Metastatic Pancreatic Cancer.
O'Reilly EM, Wainberg ZA, Wolpin BM et al.;
N Engl J Med. 2026 May 31. doi: 10.1056/NEJMoa2605555. Epub ahead of print. PMID: 42223072.
Small Bowel
Aortoenteric fistulas in the emergency setting: CT findings and diagnostic pitfalls - a pictorial review.
Carballo Cuevas E, et al.
Emerg Radiol. 2026 Apr;33(2):439-444.- Aortoenteric fistulas (AEFs) are rare but life-threatening pathologic communications between the aorta and the gastrointestinal tract, associated with extremely high mortality if not promptly recognized and treated . AEFs are classically categorized as primary or secondary. Primary AEFs arise from erosion of a native aortic aneurysm—most commonly atherosclerotic—into the adjacent bowel, typically involving the third or fourth portion of the duodenum. Secondary AEFs are more common and occur as a complication of prior aortic reconstructive surgery, usually related to graft infection, mechanical erosion, or anastomotic pseudoaneurysm formation.
Aortoenteric fistulas in the emergency setting: CT findings and diagnostic pitfalls - a pictorial review.
Carballo Cuevas E, et al.
Emerg Radiol. 2026 Apr;33(2):439-444. - The clinical presentation of aortoenteric fistulas is highly variable and frequently nonspecific. Gastrointestinal bleeding (hematemesis or melena) is the most common manifestation and may be intermittent or massive. Additional findings include abdominal or back pain, fever, sepsis, or hemorrhagic shock. Subtle early symptoms may delay recognition, and a history of prior aortic aneurysm repair or vascular intervention should raise immediate suspicion in patients presenting with gastrointestinal bleeding or unexplained sepsis.
Aortoenteric fistulas in the emergency setting: CT findings and diagnostic pitfalls - a pictorial review.
Carballo Cuevas E, et al.
Emerg Radiol. 2026 Apr;33(2):439-444. - Although few studies have separately quantified the diagnostic performance of individual CT findings, direct signs such as active contrast extravasation or visualization of graft material within the bowel lumen are considered highly specific for aortoenteric fistula, albeit infrequently observed in clinical practice. In contrast, indirect findings—including periaortic fat plane effacement, adjacent bowel wall thickening, ectopic gas, perigraft fluid, and perigraft hematoma—are more commonly detected but lack specificity and may overlap with graft infection or postoperative inflammatory changes
Aortoenteric fistulas in the emergency setting: CT findings and diagnostic pitfalls - a pictorial review.
Carballo Cuevas E, et al.
Emerg Radiol. 2026 Apr;33(2):439-444. - Etiology
Primary Aortoenteric Fistulas:
- Occur in a native aorta without a history of prior intervention
- Much more rare than secondary fistula
- Causes include:
--- Atherosclerotic Penetrating Ulcer (most common)
--- Diverticulitis
--- Foreign bodies
--- Aortitis
--- Appendicitis
--- Gastrointestinal malignancies - Etiology
Secondary Aortoenteric Fistulas
- Occur in the setting of prior surgery or intervention
- Incidence up to 0.6% in patients with prior aortic surgery or graft placement
- Thought to be secondary to prolonged pressure upon the bowel by a graft or chronic perigraft infection
- High Risk Factors:
--- Emergent surgery for a ruptured aneurysm
--- Operative complications such as reoperation or bowel injury
--- Endoleak
--- Stent Migration - Aorto Enteric Fistulae
Can occur with any portion of the gastrointestinal tract
- Classic location is the transverse portion of the duodenum (60% of cases)
- Remainder of duodenum
- Jejunum and ileum
- Stomach
- Sigmoid colon
- Ascending/descending colon - Aorto Enteric Fistulae: Primary CT Findings
- Ectopic gas either within or directly adjacent to the aortic lumen
--- Rarely, gas can be tracked from the involved bowel loop to the aorta
- Direct extravasation of contrast from the aorta into a bowel loop – extraordinarily rare
- Leakage of enteric contrast directly into the periaortic space – extraordinarily rare - Aorto Enteric Fistulae: Secondary CT Findings
- Effacement of the periaortic fat plane
- Focal thickening and tethering of a bowel loop immediately adjacent to the aorta
- Periaortic free fluid and soft tissue thickening
- Disruption of a graft or significant graft migration
- Penetrating ulcer or intramural hematoma immediately adjacent to a tethered, abnormal appearing loop of bowel - Aorto Enteric Fistulae: Mimics
- Severe perigraft infection
- Aortitis
- Mycotic aneurysms
- Perianeurysmal fibrosis
- Immediate post-operative aorta
--- Ectopic gas can be normal up to 1 month after surgery, and perigraft fluid can be normal up to 3 months after surgery
- Primary small bowel malignancies are rare, often presenting with nonspecific symptoms or as acute emergencies, which can delay diagnosis. Contrast-enhanced CT is the primary imaging modality in the emergency setting, but detection and characterization of small bowel tumors remain challenging. Cinematic rendering (CR) is a recently developed three-dimensional post-processing technique that produces photorealistic images from CT data, enhancing visualization of small bowel pathology. This pictorial review outlines the CT imaging features of major small bowel malignancies, including adenocarcinoma, carcinoid tumor, gastrointestinal stromal tumor, lymphoma, and sarcoma, and describes features that highlight the utility of CR in augmenting traditional imaging. CR offers improved visualization of mucosal abnormalities, tumor extent, vascular involvement, and textural differences, potentially increasing diagnostic confidence, supporting presurgical planning, and facilitating communication among clinicians and patients. By emphasizing the added value of CR, we aim to provide radiologists with practical guidance for identifying small bowel neoplasms and suggest that integrating advanced 3D visualization into routine CT evaluation can support timely diagnosis and management in acute care settings.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Small bowel malignant neoplasms are rare, accounting for only 0.6% of all cancers and 3% of gastrointestinal (GI) malignancies in the U.S., though incidence has risen 118% over the past four decades. Among these tumors, small bowel adenocarcinomas and carcinoid (i.e. neuroendocrine) tumors together comprise approximately 80% of cases, with each accounting for about 40%. The remaining 20–25% are made up of other malignancies, including gastrointestinal stromal tumors (GIST), lymphomas and sarcomas . Several risk factors have been identified for the development of these malignant neoplasms, such as inflammatory conditions like inflammatory bowel disease and celiac disease; hereditary syndromes including familial adenomatous polyposis, hereditary nonpolyposis colorectal cancer, and Peutz-Jeghers syndrome; and infections such as HIV. .
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Additionally, photorealistic 3D images intrinsic to CR can help increase the understanding of these pathologies for students, trainees, and patients. At our institution, cinematic rendering is performed using syngo.via software. Predefined transfer functions within the CR platform map attenuation values from the original CT dataset to specific color palettes and opacity levels, thereby controlling tissue transparency. By selecting and fine-tuning different transfer functions and adjusting clipping planes, varying degrees of translucency can be achieved to selectively emphasize specific tissues and anatomical structures, enabling the generation of high quality, clinically meaningful CR images. The authors point out that for advanced 3D postprocessing methods like CR to be used effectively in clinical practice, institutions must have high quality contrast-enhanced images, access to the specialized software, and radiologists with specific training in these technologies.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Small bowel adenocarcinoma most commonly arises in the distal duodenum or proximal jejunum through malignant transformation of glandular cells, often originating from precursor adenomas. Tumors under 2 cm are commonly overlooked, especially because bowel dilation or obstructive features and clinical symptoms are uncommon. On CT imaging, small bowel adenocarcinomas usually appear as enhancing lesions that cause irregular, circumferential or eccentric narrowing of the bowel lumen. This can lead to obstructive symptoms, and a study of 217 patients with small bowel adenocarcinoma found emergency diagnoses due to occlusion in 40% and bleeding in 24% of patients.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Carcinoid tumors of the small bowel often present with vague symptoms such as abdominal pain, but may initially be identified by carcinoid syndrome which is marked by secretory diarrhea, flushing, telangiectasia, bronchial constriction, and potential cardiac valve abnormalities secondary to bioactive substances such as serotonin secreted by carcinoid tumors associated with liver metastases. These tumors originate from chromaffin cells at the base of the crypts of Lieberkühn, most commonly in the distal ileum. They generally grow as submucosal nodules and typically demonstrate intense early hyperenhancement on imaging. Detecting these lesions can be challenging with conventional imaging techniques, given their often sub-centimeter size and submucosal location.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Desmoplastic reaction may result in angulation, tethering, and fixation of the affected small bowel loops. Mesenteric involvement and desmoplastic reaction can be clearly visualized with CR by adjusting windowing levels, allowing for detailed depiction of features such as the ‘spoke-wheel’ or stellate pattern, as well as assessment of adjacent vessels and any vascular compromise. With CR, physicians can intuitively visualize the extent and locations of multiple lesions in a single 3D view. Another finding may be carcinoid metastases, most commonly found in the liver, where their hypervascular nature leads to bright enhancement on CT and CR . Additional metastatic sites can include the lungs, bones, and peritoneum.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - “On CT imaging, GISTs are typically seen as large, lobulated, well-circumscribed, and predominantly exophytic soft tissue masses with heterogeneous enhancement patterns, although small GISTs usually appear as sharply margined, smooth-walled, homogeneous masses. While calcifications are rare, neovascularity, central necrosis, ulceration, hemorrhage, or cavitation may be observed within the masses. Consequently, CR can be particularly helpful in evaluating the internal architecture of GISTs and identifying key features, due to its capability to accentuate textural differences.”
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Primary gastrointestinal lymphoma is the most common extranodal lymphoma, most frequently affecting the ileum due to its rich lymphoid tissue, and is predominantly of B-cell origin, with only 8–10% of cases arising from T-cells. Patients can present in emergency settings due to GI perforation, occurring in about 9% of lymphoma cases, with 59% involving the small bowel, although obstruction remains atypical due to the absence of a desmoplastic reaction. Along with chronic inflammatory conditions, infections are risk factors and include Helicobacter pylori, HIV, Campylobacter jejuni and Epstein-Barr virus among others. Radiologically, key indicators of primary small bowel lymphoma include enhancing wall thickening, with adjacent multiple enlarged mesenteric lymph nodes without necrosis.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Small bowel lymphoma can be ill-defined and has various forms and can present as pseudoaneurysmal wall thickening, polypoid intraluminal masses causing intussusception, endoexoenteric cavitary masses, exophytic bulky masses invading the mesentery, or rarely, stenosing fibrotic narrowing. CR has the potential to identify and even differentiate these forms of lymphoma due to the global 3D overview it provides and its capacity to accentuate texture differences between healthy tissue and disease processes. Furthermore, the presence of associated bulky lymphadenopathy and multifocal involvement can help differentiate lymphomas from small bowel GIST and adenocarcinoma.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - Sarcomas only represent about 10% of small bowel cancers. Leiomyosarcomas are the most common, though many other subtypes can occur, each with distinct imaging features such as varying compositions . These tumors most frequently arise in the jejunum, followed by the ileum and duodenum and are typically slow growing yet aggressive . On imaging, they usually appear as large, heterogeneously enhancing masses with central necrosis and focal wall thickening and rarely tumoral calcification. Cavitation and direct communication with the bowel lumen may be observed, and their vascular nature often leads to complications such as bleeding or perforation.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617 - In summary, small bowel neoplasms are rare and often present with either nonspecific symptoms or acute emergencies, making diagnosis in the emergency setting pivotal. CECT remains the frontline imaging modality, but 3D CR offers added value by enhancing surface detail, accentuating subtle textural differences, and providing realistic depictions of tumor anatomy and surrounding structures. These advantages potentially improve detection, characterization, and assessment of vascular involvement, ultimately aiding radiologists in rapidly identifying and managing small bowel malignancies in acute care settings.
Augmenting CT evaluation of primary small bowel malignancies: The role of cinematic rendering
Zahra F. Rahmatullah · Satomi Kawamoto · Elliot K. Fishman
Emergency Radiology (2026) 33:609–617
- Behçet's disease
Behçet's disease is a chronic, systemic, relapsing vasculitis unique for its ability to involve vessels of all sizes (small, medium, and large) on both the arterial and venous sides of the circulation.
While clinical diagnosis relies on classic mucocutaneous and ocular findings, Computed Tomography (especially Multidetector CT and CT Angiography) plays a vital role in demonstrating the widespread thoracic, abdominal, and vascular manifestations that dictate prognosis - Behçet's disease: GI Findings
Behçet's frequently causes deep, penetrating ulcers typically clustering in the ileocecal region. CT findings include concentric, asymmetric bowel wall thickening, marked mucosal enhancement, and surrounding perienteric fat stranding. It can look morphologically identical to Crohn's disease, but holds a higher propensity for perforation and fistula formation. - Behçet's disease: Chest Findings
Pulmonary Artery Aneurysms (PAAs): Considered a hallmark imaging feature of the disease and a major cause of mortality.
CT Findings: Often multiple, bilateral, and saccular, frequently involving the main or lobar pulmonary arteries.
Thrombosis: Aneurysms often contain eccentric mural thrombi which can lead to total occlusion.
Venous Thrombosis: * CT Findings: High propensity for Superior Vena Cava (SVC) thrombosis, seen as a persistent filling defect with a thickened, enhancing vein wall due to active vasculitis. Marked mediastinal, chest wall, and azygous/hemiazygous collateral pathways are routinely visible.
Stomach
- Ménétrier's disease: CT Findings
Giant Cerebriform Rugal Folds: The most hallmark finding is marked, diffuse thickening of the gastric rugae. These folds project into the lumen and take on a classic "cerebriform" (brain-like) appearance. Folds can easily exceed 10" mm"to 25" mm"in width (normal folds are typically <5" mm").
Proximal Stomach Predominance: The massive wall thickening predominantly affects the fundus and body of the stomach. - Ménétrier's disease: CT Findings
Antral Sparing: Classically, the gastric antrum is relatively spared, showing normal wall thickness. While exceptions occur where the antrum is involved, striking proximal-to-distal asymmetry is highly suggestive.
Preserved Wall Stratification & Smooth Serosa: On contrast-enhanced CT, the thickened mucosa demonstrates prominent enhancement. 1 Crucially, the normal underlying wall layers (stratification) remain largely intact, and the outer serosal contour remains smooth without invasive extension into perigastric fat. - Menetrier’s disease is a rare disorder characterized by excessive gastric mucosal hypertrophy with associated protein loss. Currently, there have been fewer than 1000 reported cases of this disease. The clinical presentation is often nonspecific and can resemble signs and symptoms of gastritis. However, the protein loss can cause widespread edema, providing a clearer picture of the diagnosis of Menetrier. A detrimental complication of Menetrier’s disease is the evolution to gastric adenocarcinoma. Because of the rarity of the disease and the often vague clinical presentation, it is important for clinicians to conduct a comprehensive diagnostic workup to prevent the development of further harm to the patient.
Clinical and Radiological Features of Menetrier's Disease: A Case Report and Review of the Literature.
Keener M, et al.
Cureus. 2023 Sep 19;15(9):e45537.
- Metastases to the stomach are rare, with an incidence of 0.2– 0.7%, and primarily occur from breast, lung, esophagus, and malignant melanoma . They have a poor prognosis with resection as the preferred treatment. Melanoma is the most common source of gastric metastases with hypervascular lesions. On CT, they may appear as multiple small nodules or a solitary large mass, with or without ulceration. They are generally submucosal found in the upper and middle third of the stomach [60]. This can mimic benign subepithelial tumors or malignant lesions such as GISTs. In a patient with a known primary malignancy, subepithelial gastric lesions on CT (one or multiple) should suggest possible metastasis, particularly from tumors that often spread to the stomach, and warrant endoscopic biopsy.
From common to rare: imaging spectrum of gastric malignancies.
Arshad H, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 12. Epub ahead of print. PMID: 42118275.
- Plasmacytoma is a rare plasma cell neoplasm characterized by clonal proliferation of monoclonal plasma cells. Most lesions arise in bone (solitary bone plasmacytoma), whereas extramedullary plasmacytomas (EMPs) represent only 3–5% of plasma cell malignancies . Gastrointestinal involvement is uncommon, accounting for less than 5% of EMP cases . It originates from the submucosa or lamina propria of the gastric wall . Due to its rarity, the literature is limited to case reports and series. Clinically, it can present with gastrointestinal bleeding due to ulceration, organ-specific dysfunction due to mass effect, weight loss, and abdominal pain. These patients have a higher risk of developing multiple myeloma in the next 10 years; thus, it is important to identify it on imaging and guide decision-making for active treatment or surveillance.
From common to rare: imaging spectrum of gastric malignancies.
Arshad H, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 12. Epub ahead of print. PMID: 42118275.
- Gastric neuroendocrine tumors (G-NETs) are rare tumors that originate from the enterochromaffin-like cells of the gastric oxyntic mucosa, comprising only 1.9–2% of all neuroendocrine tumors [33]. A 50-year study of 562 G-NETs reported an increase in the incidence of G-NETs from 0.3% to 1.77%. The mean age at diagnosis is 59 years, and the 5-year survival for all stages is 81.1% .
From common to rare: imaging spectrum of gastric malignancies.
Arshad H, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 12. Epub ahead of print. PMID: 42118275. - The WHO recognizes three histologic categories: well-differentiated endocrine tumors (benign/uncertain malignancy), well-differentiated endocrine carcinomas (low-grade malignancy potential), and poorly differentiated endocrine carcinomas (high-grade malignancy potential). Well-differentiated endocrine tumors generally appear as multiple small subepithelial or polypoid tumors, while endocrine carcinomas demonstrate ulcerative tumor growth with metastasis similar to the primary tumor. The clinical symptoms are nonspecific, and carcinoid syndrome is rarely seen in gastric G-NETs. Patients can present with symptoms of GI bleeding and mass effect in case of larger tumors.
From common to rare: imaging spectrum of gastric malignancies.
Arshad H, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 12. Epub ahead of print. PMID: 42118275.
- Gastrointestinal stromal tumors (GISTs) are the most common non-epithelial mesenchymal tumors of the GI tract. They are hypothesized to originate from interstitial cells of Cajal found in the myenteric plexus . They show immunostaining for CD117 and c-KIT expression, distinguishing them histologically from leiomyoma or leiomyosarcoma . The stomach is the most common site for GISTs, accounting for 2–3% of all gastric tumors. GISTs can demonstrate varying imaging patterns that overlap with benign leiomyomas; thus, it is important to differentiate between the two for management planning. GISTs in the stomach are mostly benign, but they can be malignant, requiring appropriate chemotherapy and surgery. The most common clinical presentation is abdominal pain and GI bleeding due to ulceration of the mucosa that may lead to anemia, hematemesis, or melena.
From common to rare: imaging spectrum of gastric malignancies.
Arshad H, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 12. Epub ahead of print. PMID: 42118275. - On CT, GISTs exhibit varying sizes, growth patterns, and enhancement patterns depending on tumor aggressiveness and time of presentation. Kim et al. reported a mean size of 8.1 cm among 81 patients, ranging from 1 to 23 cm. A malignant GIST is generally a large, well-circumscribed tumor growing exophytically with a necrotic center and heterogeneous soft-tissue rim enhancement, as seen. Exophytic growth pattern can lead to displacement of the stomach and may occupy the left upper quadrant. They may show internal hemorrhage or necrosis that could lead to a large cavity containing air, air-fluid levels, or oral contrast material due to ulceration.
From common to rare: imaging spectrum of gastric malignancies.
Arshad H, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 12. Epub ahead of print. PMID: 42118275.
- Multidetector CT has improved the detection, diagnosis, and characterization of gastric malignancies, and as well as the evaluation of distant metastases. Gastric malignancies range from common types such as adenocarcinoma, lymphoma, and GISTs to rare types such as neuroendocrine tumors, liposarcoma, leiomyosarcoma, plasmacytoma, and metastases from other cancers. CT is often the preferred modality for their assessment. These tumors can appear as a single lesion, multifocal involvement, or diffuse wall thickening, and they often have overlapping imaging features, making diagnosis challenging. This pictorial review presents the CT findings of these malignancies to help radiologists understand the differential diagnosis, recognize early CT patterns, and guide appropriate management.
From common to rare: imaging spectrum of gastric malignancies.
Arshad H, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 12. Epub ahead of print. PMID: 42118275. - Water is commonly used as oral neutral contrast, however, gas producing effervescent granules can also be used. At our institution, we use water as an oral contrast. About 750–1000 ml is ingested by the patient 15–20 min before the scan and an additional 250 ml is administered just before the scan. It has been seen that CT using water as an oral contrast agent (Hydro-CT) demonstrated a 92.5% accuracy in tumor identification, comparable to endoscopy.
From common to rare: imaging spectrum of gastric malignancies.
Arshad H, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 12. Epub ahead of print. PMID: 42118275. - Adenocarcinoma is the most common gastric malignancy, constituting approximately 95% of primary gastric neoplasms [9]. Worldwide, more than 1 million people are newly diagnosed with gastric cancer each year, and it is the third leading cause of cancer-related mortality around the world. However, early gastric cancers have shown improved postoperative 5-year survival rates of more than 90%. The peak age of diagnosis is between 50 and 70 years. CT is used for both the detection and preoperative staging of gastric cancer, with an overall accuracy ranging from 69–85% for preoperative T staging; however, it reduces to 20–53% for early stages of gastric cancer.
From common to rare: imaging spectrum of gastric malignancies.
Arshad H, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 12. Epub ahead of print. PMID: 42118275. - The gastrointestinal (GI) tract is the most common location for extranodal non-Hodgkin lymphoma (NHL), accounting for 20% of the cases . In the GI tract, the stomach is the most frequent site for both primary lymphoma and systemic lymphoma with GI involvement. It is the second most common gastric malignancy, accounting for 1–7% of all gastric neoplasms, and the majority are B-cell NHL. B-cell GI lymphomas in the stomach are mostly low-grade mucosa-associated lymphoid tissue (MALT) type caused by proliferation of lymphoid tissue in the gastric mucos. They occur in the sixth decade of life, with several risk factors, including infectious agents such as Helicobacter pylori, Human immunodeficiency viruses (HIV), Epstein-Barr Virus (EBV), Hepatitis B, and Human T-lymphotropic Virus (HTLV), as well as inflammatory conditions such as celiac disease, atrophic gastritis, and parasitic infections . It presents with nonspecific clinical findings, including epigastric pain, weight loss, and anorexia.
From common to rare: imaging spectrum of gastric malignancies.
Arshad H, Chu LC, Fishman EK.
Abdom Radiol (NY). 2026 May 12. Epub ahead of print. PMID: 42118275.
Vascular
- Purpose: To quantify postimplementation concordance between a U.S. Food and Drug Administration–cleared artificial intelligence (AI) tool and AI-informed radiologists for pulmonary embolism (PE) detection at CT pulmonary angiography, with real-time adjudication of discordances.
Materials and Methods: A commercial PE AI tool was retrospectively implemented in the clinic across an integrated network (August 9, 2021 February 20, 2023). Adult CT pulmonary angiographic acquisitions underwent real-time AI analysis and radiologist interpretation. Radiologist-AI disagreements triggered adjudication by thoracic radiologists via the AI quality oversight process. Adjudicator diagnosis served as the reference standard for discordant cases. Concordance was measured and diagnostic performance of radiologists and AI was compared using adjudication for discordant cases.
Conclusion: In large-scale deployment, AI showed high concordance with radiologists and made meaningful contributions in discordant reviews while expert oversight confirmed complementary roles and highlighted scenarios of radiologist-AI divergence.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017 - Results: A total of 32 501 CT pulmonary angiographic acquisitions obtained from 29 492 patients (mean age, 62.4 years Å} 18.6 [SD], 17 424 female patients) were evaluated. PE positivity was 9.93% (3226 of 32 501). Overall concordance was 97.79% (95% CI: 97.62, 97.94) and was higher for AI-negative than for AI-positive examinations (98.18% vs 93.75%; P < .001). Expert adjudication favored the radiologist in 88.73% of discordances. The rate of unique diagnosis by the interpreting radiologist (483 of 3226 [14.97%]) was approximately 19 times that of the AI tool alone (26 of 3226 [0.81%]). Concordance varied by PE features: acute versus chronic (87.34% vs 60.12%; P < .001) and location (central, 95.79%; lobar and/or segmental, 83.81%; subsegmental, 58.62%; all P < .001).
Conclusion: In large-scale deployment, AI showed high concordance with radiologists and made meaningful contributions in discordant reviews while expert oversight confirmed complementary roles and highlighted scenarios of radiologist-AI divergence.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017 - ■ In a retrospective study of 29 492 patients, radiologist–artificial intelligence (AI) concordance for pulmonary embolism (PE) detection was high (97.79%) and greater for AI-negative than for AI positive examinations (98.18% vs 93.75%), suggesting AI-negative outputs provide supportive signal while AI-positive alerts merit scrutiny.
■ Expert adjudication favored the radiologist in 88.73% of discordances,and AI contributed selectively, underscoring complementary roles in radiologist-led workflows.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017 - These findings support complementary roles for radiologists and AI, with radiologist-led oversight remaining essential. Radiologists should avoid overreliance on AI, remain vigilant in search patterns, and scrutinize AI-positive results. The clinical value of the AI tool is maximized under discerning, radiologist-led oversight, reinforcing AI’s role as diagnostic support and triage rather than a standalone solution. AI should be directly supervised by radiologists who render final diagnoses and should not be used in alternative workflows without imaging-expert oversight.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017 - In conclusion, we observed a high ~98% agreement between radiologists and the AI tool for the diagnosis of PE, with both AI and AI-informed radiologists offering unique diagnostic contributions. Future work will examine the attributes of radiologist- AI disagreement cases to determine which imaging, patient, or report factors are more likely to confound humanreaders compared with AI. We will also quantify the triage benefits of this tool, including its impact on turnaround time for PE-positive cases.
Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations
Shlomit Goldberg-Stein et al.
Radiology: Artificial Intelligence 2026; 8(4):e250017
Aortoenteric fistulas in the emergency setting: CT findings and diagnostic pitfalls - a pictorial review.
Carballo Cuevas E, et al.
Emerg Radiol. 2026 Apr;33(2):439-444.- Aortoenteric fistulas (AEFs) are rare but life-threatening pathologic communications between the aorta and the gastrointestinal tract, associated with extremely high mortality if not promptly recognized and treated . AEFs are classically categorized as primary or secondary. Primary AEFs arise from erosion of a native aortic aneurysm—most commonly atherosclerotic—into the adjacent bowel, typically involving the third or fourth portion of the duodenum. Secondary AEFs are more common and occur as a complication of prior aortic reconstructive surgery, usually related to graft infection, mechanical erosion, or anastomotic pseudoaneurysm formation.
Aortoenteric fistulas in the emergency setting: CT findings and diagnostic pitfalls - a pictorial review.
Carballo Cuevas E, et al.
Emerg Radiol. 2026 Apr;33(2):439-444. - The clinical presentation of aortoenteric fistulas is highly variable and frequently nonspecific. Gastrointestinal bleeding (hematemesis or melena) is the most common manifestation and may be intermittent or massive. Additional findings include abdominal or back pain, fever, sepsis, or hemorrhagic shock. Subtle early symptoms may delay recognition, and a history of prior aortic aneurysm repair or vascular intervention should raise immediate suspicion in patients presenting with gastrointestinal bleeding or unexplained sepsis.
Aortoenteric fistulas in the emergency setting: CT findings and diagnostic pitfalls - a pictorial review.
Carballo Cuevas E, et al.
Emerg Radiol. 2026 Apr;33(2):439-444. - Although few studies have separately quantified the diagnostic performance of individual CT findings, direct signs such as active contrast extravasation or visualization of graft material within the bowel lumen are considered highly specific for aortoenteric fistula, albeit infrequently observed in clinical practice. In contrast, indirect findings—including periaortic fat plane effacement, adjacent bowel wall thickening, ectopic gas, perigraft fluid, and perigraft hematoma—are more commonly detected but lack specificity and may overlap with graft infection or postoperative inflammatory changes
Aortoenteric fistulas in the emergency setting: CT findings and diagnostic pitfalls - a pictorial review.
Carballo Cuevas E, et al.
Emerg Radiol. 2026 Apr;33(2):439-444. - Etiology
Primary Aortoenteric Fistulas:
- Occur in a native aorta without a history of prior intervention
- Much more rare than secondary fistula
- Causes include:
--- Atherosclerotic Penetrating Ulcer (most common)
--- Diverticulitis
--- Foreign bodies
--- Aortitis
--- Appendicitis
--- Gastrointestinal malignancies - Etiology
Secondary Aortoenteric Fistulas
- Occur in the setting of prior surgery or intervention
- Incidence up to 0.6% in patients with prior aortic surgery or graft placement
- Thought to be secondary to prolonged pressure upon the bowel by a graft or chronic perigraft infection
- High Risk Factors:
--- Emergent surgery for a ruptured aneurysm
--- Operative complications such as reoperation or bowel injury
--- Endoleak
--- Stent Migration - Aorto Enteric Fistulae
Can occur with any portion of the gastrointestinal tract
- Classic location is the transverse portion of the duodenum (60% of cases)
- Remainder of duodenum
- Jejunum and ileum
- Stomach
- Sigmoid colon
- Ascending/descending colon - Aorto Enteric Fistulae: Primary CT Findings
- Ectopic gas either within or directly adjacent to the aortic lumen
--- Rarely, gas can be tracked from the involved bowel loop to the aorta
- Direct extravasation of contrast from the aorta into a bowel loop – extraordinarily rare
- Leakage of enteric contrast directly into the periaortic space – extraordinarily rare - Aorto Enteric Fistulae: Secondary CT Findings
- Effacement of the periaortic fat plane
- Focal thickening and tethering of a bowel loop immediately adjacent to the aorta
- Periaortic free fluid and soft tissue thickening
- Disruption of a graft or significant graft migration
- Penetrating ulcer or intramural hematoma immediately adjacent to a tethered, abnormal appearing loop of bowel - Aorto Enteric Fistulae: Mimics
- Severe perigraft infection
- Aortitis
- Mycotic aneurysms
- Perianeurysmal fibrosis
- Immediate post-operative aorta
--- Ectopic gas can be normal up to 1 month after surgery, and perigraft fluid can be normal up to 3 months after surgery