google ads
Search

Everything you need to know about Computed Tomography (CT) & CT Scanning

Deep Learning: Deep Learning and the Pancreas Imaging Pearls - Educational Tools | CT Scanning | CT Imaging | CT Scan Protocols - CTisus
Imaging Pearls ❯ Deep Learning ❯ Deep Learning and the Pancreas

-- OR --

  • BACKGROUND & AIMS: Our purpose was to detect pancreatic ductal adenocarcinoma(PDAC) at the prediagnostic stage (3-36 months prior to clinical diagnosis) using radiomics based machine learning (ML) models, and to compare performance against radiologists in a case control study.
    METHODS: Volumetric pancreas segmentation was performed on prediagnostic CTs (median interval between CT and PDAC diagnosis: 398 days) of 155 patients and an age-matched cohort of 265 subjects with normal pancreas. Total 88 first order and gray level radiomic features were extracted and 34 features were selected through LASSO-based feature selection method. Dataset was randomly divided into training (292 CTs: 110 prediagnostic and 182 controls) and test subsets (128 CTs: 45 prediagnostic and 83 controls). Four ML classifiers - K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Random Forest (RF) and XGBoost (XGB) - were evaluated. Specificity of model with highest accuracy was further validated on an independent internal dataset (n=176) and the public NIH dataset (n=80). Two radiologists (R4 and R5) independently evaluated the pancreas on a five-point diagnostic scale.
    Radiomics-Based Machine Learning Models Can Detect Pancreatic Cancer on Prediagnostic CTs at a Substantial Lead Time Prior to Clinical Diagnosis
    Sovanlal Mukherjee,et al.
    Gastroenterology (2022),doi:https://doi.org/10.1053/j.gastro.2022.06.066.
  • “Our study has limitations. The retrospective nature of the study is generally prone to selection bias. As with other radiomics studies, the precise pathologic correlates of the radiomic features that constitute the ML classifiers are not entirely known. We did not investigate the impact of differences in all the acquisition or post-processing parameters (e.g., voxel width, bin width, etc.) on the classifiers, which will be subject of the next phase of our ongoing investigation. Although we validated the high specificity of the SVM classifier on an independent internal cohort of control CTs as well as on the public NIH-PCT dataset, the sample size of these cohorts was small and the subjects in these cohorts were relatively younger. Thus, prospective larger cohorts with both cases and controls are warranted for further validation. Such prospective studies would also help determine the optimal operating point for the models to avoid a high false positive rate in context of a screening paradigm.”  
    Radiomics-Based Machine Learning Models Can Detect Pancreatic Cancer on Prediagnostic CTs at a Substantial Lead Time Prior to Clinical Diagnosis
    Sovanlal Mukherjee,et al.
    Gastroenterology (2022),doi:https://doi.org/10.1053/j.gastro.2022.06.066.
  • “In conclusion, we detected and quantified the imaging signature of early pancreatic  carcinogenesis from volumetrically segmented normal pancreas on standard-of-care CTs. The radiomics-based ML classifiers had high discrimination accuracy for classification of pancreas into prediagnostic for PDAC versus normal. The high accuracy of the SVM model was validated on CTs from external institutions. Its high specificity was generalizable on an independent internal cohort and on an external public dataset. In contrast, radiologist readers had low interreader agreement, sensitivity, and discrimination accuracy, which shows that novel AI-based approaches can detect PDAC at a subclinical stage when it is beyond the scope of the human interrogation. Prospective validation of these ML models and their integration with complementary blood and other fluid-based biomarkers has the potential to further improve cancer prediction capabilities at the prediagnostic or symptom-free stage. Such models also have the potential to elucidate the longitudinal changes of carcinogenesis that precede the clinical diagnosis of PDAC. Finally, such models can be deployed to detect early cancer in ongoing clinical trials such as the Early Detection Initiative that seeks to evaluate outcomes of a screening strategy utilizing clinical risk-prediction models and CT in cohorts at high-risk for PDAC.”
    Radiomics-Based Machine Learning Models Can Detect Pancreatic Cancer on Prediagnostic CTs at a Substantial Lead Time Prior to Clinical Diagnosis
    Sovanlal Mukherjee,et al.
    Gastroenterology (2022),doi:https://doi.org/10.1053/j.gastro.2022.06.066.
  • Objective: Quality gaps in medical imaging datasets lead to profound errors in experiments. Our objective was to characterize such quality gaps in public pancreas imaging datasets (PPIDs), to evaluate their impact on previously published studies, and to provide post-hoc labels and segmentations as a value-add for these PPIDs.
    Conclusion: Substantial quality gaps, sources of bias, and high proportion of CTs unsuitable for AI characterize the available limited PPIDs. Published studies on these PPIDs do not account for these quality gaps. We complement these PPIDs through post-hoc labels and segmentations for public release on the TCIA portal. Collaborative efforts leading to large, well-curated PPIDs supported by adequate documentation are critically needed to translate the promise of AI to clinical practice.
    Quality gaps in public pancreas imaging datasets: Implications & challenges for AI applications
    Garima Suman et al.
    Pancreatology 21 (2021) 1001-1008
  • "Public medical imaging datasets have stimulated widespread interest to explore AI to address unmet healthcare needs. In order to fully leverage these public datasets, there is a critical need to understand their strengths and limitations. Our study of public datasets in the pancreas imaging domain identified only three public datasets. The MSD dataset is the largest one with 420 CTs. Both the NIH-PCT and the TCIA PDA datasets have less than 100 CTs each. These datasets are insufficient for deep learning applications, which require very large volumes of data. There is a general hesitation to share digital assets due to concerns related to data ownership and patient privacy. Ongoing developments in federated learning architecture and privacy-preserving AI could promote wider sharing of such datasets.”
    Quality gaps in public pancreas imaging datasets: Implications & challenges for AI applications
    Garima Suman et al.
    Pancreatology 21 (2021) 1001-1008
  • “Finally, presence of medical devices such as stents is another critical confounding factor. In the context of PDA, a tumor classification, or detection model can learn to associate the presence of a biliary stent with the diagnosis of PDA, which can lead to inadvertent overestimation of the model’s performance. Secondly, the course of such stents through the pancreatic head results in streak, artifacts and can obscure delineation of tumors in the pancreatic head. These challenges can increase the variability in tumor segmentation or result in the stent being included in segmentation mask with consequent errors in AI models. Therefore, if CTs with stents form a part of PPIDs, these should be explicitly specified in the metadata to ensure that users can make an Informed decision regarding their potential use for AI experiments.”
    Quality gaps in public pancreas imaging datasets: Implications & challenges for AI applications
    Garima Suman et al.
    Pancreatology 21 (2021) 1001-1008
  • "In  summary, there is a need for carefully curated public imaging datasets supported by adequate documentation in the pancreas imaging domain. The available datasets for pancreatic pathologies have substantial quality gaps, sources of bias, and high proportion of CTs unsuitable for AI experiments. In our assessment, the factors  responsible for such quality gaps include general hesitation to share highly curated digital assets due to concerns related to data ownership and patient privacy, absence of tangible incentives fordata sharing, limited guidance on the dataset preparation process, inadequate involvement of domain experts in dataset curation process, and lack of awareness of the impact of insufficient documentation on the AI development pipeline.”
    Quality gaps in public pancreas imaging datasets: Implications & challenges for AI applications
    Garima Suman et al.
    Pancreatology 21 (2021) 1001-1008
  • BACKGROUND & AIMS: Our purpose was to detect pancreatic ductal adenocarcinoma (PDAC) at the prediagnostic stage (3-36 months prior to clinical diagnosis) using radiomics based machine learning (ML) models, and to compare performance against radiologists in a case control study.  
    CONCLUSIONS: Radiomics-based ML models can detect PDAC from normal pancreas when it is beyond human interrogation capability at a substantial lead time prior to clinical diagnosis. Journal Pre-proof 6 Prospective validation and integration of such models with complementary fluid-based biomarkers has the potential for PDAC detection at a stage when surgical cure is a possibility
    Radiomics-Based Machine Learning Models Can Detect Pancreatic Cancer on Prediagnostic CTs at a Substantial Lead Time Prior to Clinical Diagnosis,
    Sovanlal Mukherjee et al.
    Gastroenterology 2022 (in press)
  • METHODS
    Volumetric pancreas segmentation was performed on prediagnostic CTs (median interval between CT and PDAC diagnosis: 398 days) of 155 patients and an age-matched cohort of 265 subjects with normal pancreas. Total 88 first order and gray level radiomic features were extracted and 34 features were selected through LASSO-based feature selection method. Dataset was randomly divided into training (292 CTs: 110 prediagnostic and 182 controls) and test subsets (128 CTs: 45 prediagnostic and 83 controls). Four ML classifiers - K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Random Forest (RF) and XGBoost (XGB) - were evaluated. Specificity of model with highest accuracy was further validated on an independent internal dataset (n=176) and the public NIH dataset (n=80). Two radiologists (R4 and R5) independently evaluated the pancreas on a five-point diagnostic scale.
  •  RESULTS
    Median (range) time between prediagnostic CTs of the test subset and PDAC diagnosis was 386 (97-1092) days. SVM had the highest sensitivity (mean; 95% CI) (95.5; 85.5-100.0), specificity (90.3; 84.3-91.5), F1-score (89.5; 82.3-91.7), AUC (0.98; 0.94-0.98) and accuracy (92.2%; 86.7-93.7) for classification of CTs into prediagnostic versus normal. All three other ML models KNN, RF, and XGB had comparable AUCs (0.95, 0.95, and 0.96, respectively). The high specificity of SVM was generalizable to both the independent internal (92.6%) and the NIH dataset (96.2%). In contrast, inter-reader radiologist agreement was only fair (Cohen’s kappa 0.3) and their mean AUC (0.66; 0.46-0.86) was lower than each of the four ML models (AUCs: 0.95-0.98) (p < 0.001). Radiologists also recorded false positive indirect findings of PDAC in control subjects (n=83) (7% R4, 18% R5). 
  • CONCLUSIONS
    Radiomics-based ML models can detect PDAC from normal pancreas when it is beyond human interrogation capability at a substantial lead time prior to clinical diagnosis. Prospective validation and integration of such models with complementary fluid-based biomarkers has the potential for PDAC detection at a stage when surgical cure is a possibility. 
  • “These observations support the biologic insights from prior studies that the prediagnostic stage of PDAC is marked by substantial cellular activity and infiltration, which results in marked tissue heterogeneity . Our study suggests that this tissue heterogeneity is beyond the human perceptive ability but can be captured and leveraged for actionable insights through computational postprocessing techniques such as radiomics.”  
    Radiomics-Based Machine Learning Models Can Detect Pancreatic Cancer on Prediagnostic CTs at a Substantial Lead Time Prior to Clinical Diagnosis,
    Sovanlal Mukherjee et al.
    Gastroenterology 2022 (in press)
  • “The radiomics-based ML classifiers had high discrimination accuracy for classification of pancreas into prediagnostic for PDAC versus normal. The high accuracy of the SVM model was validated on CTs from external institutions. Its high specificity was generalizable on an independent internal cohort and on an external public dataset. In contrast, radiologist readers had low interreader agreement, sensitivity, and discrimination accuracy, which shows that novel AI-based approaches can detect PDAC at a subclinical stage when it is beyond the scope of the human interrogation. Prospective validation of these ML models and their integration with complementary blood and other fluid-based biomarkers has the potential to further improve cancer prediction capabilities at the prediagnostic or symptom-free stage. Such models also have the potential to elucidate the longitudinal changes of carcinogenesis that precede the clinical diagnosis of PDAC.”
    Radiomics-Based Machine Learning Models Can Detect Pancreatic Cancer on Prediagnostic CTs at a Substantial Lead Time Prior to Clinical Diagnosis,
    Sovanlal Mukherjee et al.
    Gastroenterology 2022 (in press)
  • Background The diagnostic performance of CT for pancreatic cancer is interpreter-dependent, and approximately 40% of tumours smaller than 2 cm evade detection. Convolutional neural networks (CNNs) have shown promise in image analysis, but the networks’ potential for pancreatic cancer detection and diagnosis is unclear. We aimed to investigate whether CNN could distinguish individuals with and without pancreatic cancer on CT, compared with radiologist interpretation.
    Interpretation CNN could accurately distinguish pancreatic cancer on CT, with acceptable generalisability to images of patients from various races and ethnicities. CNN could supplement radiologist interpretation.
    Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation
    Kao-Lang Liu et al.
    Lancet Digital Health 2020; 2: e303–13
  • “The promise of artificial intelligence (AI) to improve and reduce inequities in access, quality, and appropriateness of high-quality diagnosis remains largely unfulfilled. Vast clinical data sets, extensive computational capacity, and highly developed and accessible machine learning tools have resulted in numerous publications that describe high-performing algorithmic approachesfor a variety of diagnostic tasks. However, such approaches remain largely unadopted in clinical practice.”
    Rethinking Algorithm Performance Metrics for Artificial Intelligence in Diagnostic Medicine
    Reyna MA et al.
    JAMA Published online July 8, 2022 
  • Methods In this retrospective, diagnostic study, contrast-enhanced CT images of 370 patients with pancreatic cancer and 320 controls from a Taiwanese centre were manually labelled and randomly divided for training and validation (295 patients with pancreatic cancer and 256 controls) and testing (75 patients with pancreatic cancer and 64 controls; local test set 1). Images were preprocessed into patches, and a CNN was trained to classify patches as cancerous or non-cancerous. Individuals were classified as with or without pancreatic cancer on the basis of the proportion of patches diagnosed as cancerous by the CNN, using a cutoff determined using the training and validation set. The CNN was further tested with another local test set (101 patients with pancreatic cancers and 88 controls; local test set 2) and a US dataset (281 pancreatic cancers and 82 controls). Radiologist reports of pancreatic cancer images in the local test sets were retrieved for comparison.
    Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation
    Kao-Lang Liu et al.
    Lancet Digital Health 2020; 2: e303–13
  • Findings Between Jan 1, 2006, and Dec 31, 2018, we obtained CT images. In local test set 1, CNN-based analysis hada sensitivity of 0·973, specificity of 1·000, and accuracy of 0·986 (area under the curve [AUC] 0·997 (95% CI 0·992–1·000). In local test set 2, CNN-based analysis had a sensitivity of 0·990, specificity of 0·989, and accuracy of 0·989 (AUC 0·999 [0·998–1·000]). In the US test set, CNN-based analysis had a sensitivity of 0·790, specificity of 0·976, and accuracy of 0·832 (AUC 0·920 [0·891–0·948)]. CNN-based analysis achieved higher sensitivity than radiologists did (0·983 vs 0·929, difference 0·054 [95% CI 0·011–0·098]; p=0·014) in the two local test sets combined. CNN missed three (1·7%) of 176 pancreatic cancers (1·1–1·2 cm). Radiologists missed 12 (7%) of 168 pancreatic cancers (1·0–3·3 cm), of which 11 (92%) were correctly classified using CNN. The sensitivity of CNN for tumours smaller than 2 cm was 92·1% in the local test sets and 63·1% in the US test set.
    Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation
    Kao-Lang Liu et al.
    Lancet Digital Health 2020; 2: e303–13
  • Added value of this study We trained a CNN using contrast enhanced-CT images of Asian patients to distinguish pancreatic cancer from healthy pancreases. CNN achieved excellent accuracy and improved sensitivity compared with radiologist interpretation in independent Asian test sets, with acceptable performance in a North American test set obtained from patients of various races and ethnicities using diverse scanners and settings. These results provide the first solid proof of concept that CNN can capture the elusive CT features of pancreatic cancer to assist and supplement radiologists in the detection and diagnosis of pancreatic cancer.
    Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation
    Kao-Lang Liu et al.
    Lancet Digital Health 2020; 2: e303–13
  • “Implications of all the available evidence CNN can accurately differentiate pancreatic cancer from non-cancerous pancreas, and with improvements mightaccommodate variations in patient race and ethnicity and imaging parameters that are inevitable in real-world clinical practice. CNN holds promise for developing computer-aided detection and diagnosis tools for pancreatic cancer to supplement radiologist interpretation.”
    Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation
    Kao-Lang Liu et al.
    Lancet Digital Health 2020; 2: e303–13
  • "Artificial intelligence (AI) can speed up pancreatic cancer identification, boost tumor clearance, and detect recurrent tumors during postoperative surveillance as a feasible treatment. The capabilities of AI in identifying tumor resectability, differentiating between borderline and locally progressed pancreatic ductal adenocarcinoma (PDAC), and calculating pancreatic fatty infiltration should be further investigated and developed in the future study.”
    Pancreatic Cancer Detection using Machine and Deep Learning Techniques  
    Gupta A et al.
    2022 2nd International Conference on Innovative Practices in Technology and Management (ICIPTM)
  • "According to the findings of this study, there is a lot of interest in using machine and deep learning algorithms to predict pancreatic cancer progression. When compared to other techniques, machine learning outperformed well for various datasets. As demonstrated in figure 3, The Bayesian model produced the most significant auc of 0.94, the genetic algorithm had the best sensitivity and specificity for detecting the pancreatic tumor, with 96.7% and 82.5%, respectively.”
    Pancreatic Cancer Detection using Machine and Deep Learning Techniques  
    Gupta A et al.
    2022 2nd International Conference on Innovative Practices in Technology and Management (ICIPTM)
  • “There’s been a lot of recent AI work involving dermatology and using it for melanoma screening. Will primary care physicians augmented with AI reach a point where patients don’t need to be referred to a dermatologist? Or will dermatologists be the ones using augmented AI models to help them identify which moles need a biopsy? This survey data, which indicates that few dermatologists are using AI compared to more widespread use in primary care, suggests that, at least in some fields, early AI use cases are being used in primary care settings and that there may be more value here.”  
    Growing Use and Confidence in Artificial Intelligence for Care Delivery  
    Gordon W
    NEJM Catalyst Innovations in Care Delivery 2022; 04  10.1056/CAT.22.0095  Vol. 3 No. 4 | April 2022
  • IPMN and Cystic Pancreatic Lesions: Guidelines
    - Do people actually follow the guidelines?
    - Are they numerous interpretations of the guidelines?
    - Do the guidelines actually work and whats its success?
    - Is there a better way than current practice?
  • AI and Pancreatic Cystic Lesions
    - Radiomics
    - Deep Learning algorithmns
    - Combine Radiomics and Deep Learning algorithmns
  • Background: Society consensus guidelines are commonly used to guide management of pancreatic cystic neoplasms( PCNs). However, downsides of these guidelines include unnecessary surgery and missed malignancy. The aim of this study was to use computed tomography (CT)-guided deep learning techniques to predict malignancy of PCNs.
    Materials and Methods: Patients with PCNs who underwent resection were retrospectively reviewed. Axial images of the mucinous cystic neoplasms were collected and based on final pathology were assigned a binary outcome of advanced neoplasia or benign. Advanced neoplasia was defined as adenocarcinoma or intraductal papillary mucinous neoplasm with high-grade dysplasia. A convolutional neural network (CNN) deep learning model was trained on 66% of images, and this trained model was used to test 33% of images. Predictions from the deep learning model were compared to Fukuoka guidelines.
    Results: Twenty-seven patients met the inclusion criteria, with 18 used for training and 9 for model testing. The trained deep learning model correctly predicted 3 of 3 malignant lesions and 5 of 6 benign lesions. Fukuoka guidelines correctly classified 2 of 3 malignant lesions as high risk and 4 of 6 benign lesions as worrisome. Following deep learning model predictions would have avoided 1 missed malignancy and 1 unnecessary operation.
    Discussion: In this pilot study, a deep learning model correctly classified 8 of 9 PCNs and performed better than consensus guidelines. Deep learning can be used to predict malignancy of PCNs; however, further model improvements are necessary before clinical use.
    Use of Artificial Intelligence Deep Learning to Determine the Malignant Potential of Pancreatic Cystic Neoplasms With Preoperative Computed Tomography Imaging
    Michael D. Watson et al.
    The American Surgeon 2021, Vol. 87(4) 602–607
  • “The guidelines describe high-risk and worrisome criteria of PCNs based on preoperative imaging and clinical symptoms. Retrospective studies have demonstrate high specificity; however, these guidelines havepoor sensitivity, and using guidelines alone would haveled to missing approximately 50% of advanced neoplasia in 1 series. This results in nearly 20% of patients with a benign lesion undergoing operations with high morbidity. Given that existing guidelines are insufficient for identification of high-risk features of PCNs, an opportunity exists to leverage innovative new technologies toaid in diagnosis of potentially malignant lesions.”
    Use of Artificial Intelligence Deep Learning to Determine the Malignant Potential of Pancreatic Cystic Neoplasms With Preoperative Computed Tomography Imaging
    Michael D. Watson et al.
    The American Surgeon 2021, Vol. 87(4) 602–607
  • “In conclusion, this pilot study with a small group ofpatients (n = 27) demonstrated that a deep learning model based only on preoperative CT imaging was better able to predict advanced neoplasia than generally accepted  consensus guidelines. Further study should be directed toward creation of a larger, more sophisticated model and ultimately prospective validation of a model for predictionof malignant behavior of PCNs.”
    Use of Artificial Intelligence Deep Learning to Determine the Malignant Potential of Pancreatic Cystic Neoplasms With Preoperative Computed Tomography Imaging
    Michael D. Watson et al.
    The American Surgeon 2021, Vol. 87(4) 602–607
  • “In this pilot study for prediction of advanced neoplasia from preoperative CT imaging, a deep learning model was superior to consensus guidelines. The deep learning model was able to correctly classify 8 of 9 high-risk PCNs,while consensus guidelines were correct 6 of 9 times. In terms of possible patient outcomes, following predictions from the deep learning model would have resulted in 1unnecessary operation, while following consensusguidelines would have resulted in 2 unnecessary operationsand 1 missed neoplasm.”
    Use of Artificial Intelligence Deep Learning to Determine the Malignant Potential of Pancreatic Cystic Neoplasms With Preoperative Computed Tomography Imaging
    Michael D. Watson et al.
    The American Surgeon 2021, Vol. 87(4) 602–607
  • OBJECTIVE: To stratify high risk individuals for PDAC by identifying predictive features in pre-diagnostic abdominal Computed Tomography (CT) scans.
    METHODS: A set of CT features, potentially predictive of PDAC, was identified in the analysis of 4000 raw radiomic parameters extracted from pancreases in pre-diagnostic scans. The naïve Bayes classifier was then developed for automatic classification of CT scans of the pancreas with high risk for PDAC. A set of 108 retrospective CT scans (36 scans from each healthy control, pre-diagnostic, and diagnostic group) from 72 subjects was used for the study. Model development was performed on 66 multiphase CT scans, whereas external validation was performed on 42 venous-phase CT scans.
    RESULTS: The system achieved an average classification accuracy of 86% on the external dataset.
    CONCLUSIONS: Radiomic analysis of abdominal CT scans can unveil, quantify, and interpret micro-level changes in the pre-diagnostic pancreas and can efficiently assist in the stratification of high risk individuals for PDAC.  
    Predicting pancreatic ductal adenocarcinoma using artificial intelligence analysis of pre-diagnostic computed tomography images  
    Qureshia TA et al.
    Cancer Biomarkers 33 (2022) 211–217 
  • OBJECTIVE: To stratify high risk individuals for PDAC by identifying predictive features in pre-diagnostic abdominal Computed Tomography (CT) scans.
    CONCLUSIONS: Radiomic analysis of abdominal CT scans can unveil, quantify, and interpret micro-level changes in the pre-diagnostic pancreas and can efficiently assist in the stratification of high risk individuals for PDAC.  
    Predicting pancreatic ductal adenocarcinoma using artificial intelligence analysis of pre-diagnostic computed tomography images  
    Qureshia TA et al.
    Cancer Biomarkers 33 (2022) 211–217 
  • “In this study, we identified unique features in pre- diagnostic CT scans that are not appreciated by human eyes but are potentially predictive of PDAC and developed a classifier that performed PDAC prediction by automatically identifying pre-diagnostic scans when mixed with healthy control scans. The proposed model is highly stable, and results are satisfactory, encoring researchers to replicate the model for further validation on large dataset.”
    Predicting pancreatic ductal adenocarcinoma using artificial intelligence analysis of pre-diagnostic computed tomography images  
    Qureshia TA et al.
    Cancer Biomarkers 33 (2022) 211–217 
  • – Diagnostic: A CT scan of a patient with histopatho- logically confirmed PDAC and visible tumor on a CT scan. Patients with any type of pancreatectomy were excluded from this group.  
    – Pre-diagnostic: A CT scan from the same patient (as in the Diagnostic group), acquired between 6 months to 3 years prior to PDAC diagnosis, when no sign of PDAC or tumor was present.  
    – Healthy control: A contrast-enhanced abdominal CT scan of a different subject whose pancreas was healthy. The age and gender of each subject in the healthy control group and the year of their scan were matched to those of exactly one unique pa- tient in the pre-diagnostic group to limit morpho- logical and instrumentation variabilities, respec- tively. 
  • “Although the data repositories of both CSMC and KPMC were explored exhaustively, the amount of eligible data found was low as the pre-diagnostic scans are rarely available. Analysis on a limited dataset might suffer an overfitting problem. However, the purpose of the current study was to have proof of the concept and to encourage researchers to establish a large dataset with a collaboration for extensive training and validation of the model. A large dataset will also allow performing a biological interpretation of predictors and forming their correlation with genetic heterogeneity. A rigorous model can be a supporting tool in prospective studies and will help to increase the rate of diagnosis at an early stage.”
    Predicting pancreatic ductal adenocarcinoma using artificial intelligence analysis of pre-diagnostic computed tomography images  
    Qureshia TA et al.
    Cancer Biomarkers 33 (2022) 211–217 
  • The latest advances in AI in gastroenterology and hepatology are promising for aspect many fields of clinical care, from detection of neoplastic lesions on endoscopic assessment and improving current survival models to predicting treatment response. The application of AI to large and complex datasets may assist in the identification of new associations between variables, potentially leading to changes in clinical practice. Furthermore, the use of AI-assisted technologies has the potential to dramatically improve the quality of care. Finally, the time for assisted precision medicine is at hand, with the AI being able to tailor a treatment regimen or potentially predict the response to treatment in a specific patient based on extensive amounts of clinical data from large patient datasets. It is important to realize that, while AI currently does not substitute human clinical reasoning, it has a bright future in the betterment of patient care.
    Artificial intelligence in gastroenterology: A state-of-the-art review  
    Kröner, Paul T et al.  
    World journal of gastroenterology vol. 27,40 (2021): 6794-6824
  • Purpose: Current diagnostic and treatment modalities for pancreatic cysts (PCs) are invasive and are associated with patient morbidity. The purpose of this study is to develop and evaluate machine learning algorithms to delineate mucinous from non-mucinous PCs using non-invasive CT-based radiomics.
    Methods: A retrospective, single-institution analysis of patients with non-pseudocystic PCs, contrast-enhanced computed tomography scans within 1 year of resection, and available surgical pathology were included. A quantitative imaging software platform was used to extract radiomics. An extreme gradient boosting (XGBoost) machine learning algorithm was used to create mucinous classifiers using texture features only, or radiomic/radiologic and clinical combined models. Classifiers were compared using performance scoring metrics. Shapely additive explanation (SHAP) analyses were conducted to identify variables most important in model construction.  
    Results: Overall, 99 patients and 103 PCs were included in the analyses. Eighty (78%) patients had mucinous PCs on surgical pathology. Using multiple fivefold cross validations, the texture features only and combined XGBoost mucinous classifiers demonstrated an area under the curve of 0.72 ± 0.14 and 0.73 ± 0.14, respectively. By SHAP analysis, root mean square, mean attenuation, and kurtosis were the most predictive features in the texture features only model. Root mean square, cyst location, and mean attenuation were the most predictive features in the combined model.  
    Conclusion: Machine learning principles can be applied to PC texture features to create a mucinous phenotype classifier. Model performance did not improve with the combined model. However, specific radiomic, radiologic, and clinical features most predictive in our models can be identified using SHAP analysis. 
  • Purpose: Current diagnostic and treatment modalities for pancreatic cysts (PCs) are invasive and are associated with patient morbidity. The purpose of this study is to develop and evaluate machine learning algorithms to delineate mucinous from non-mucinous PCs using non-invasive CT-based radiomics.
    Methods: A retrospective, single-institution analysis of patients with non-pseudocystic PCs, contrast-enhanced computed tomography scans within 1 year of resection, and available surgical pathology were included. A quantitative imaging software platform was used to extract radiomics. An extreme gradient boosting (XGBoost) machine learning algorithm was used to create mucinous classifiers using texture features only, or radiomic/radiologic and clinical combined models. Classifiers were compared using performance scoring metrics. Shapely additive explanation (SHAP) analyses were conducted to identify variables most important in model construction.  
    Machine learning principles applied to CT radiomics to predict mucinous pancreatic cysts  
    Adam M. Awe et al.
    Abdominal Radiology (2022) 47:221–231 
  • Results: Overall, 99 patients and 103 PCs were included in the analyses. Eighty (78%) patients had mucinous PCs on surgical pathology. Using multiple fivefold cross validations, the texture features only and combined XGBoost mucinous classifiers demonstrated an area under the curve of 0.72 ± 0.14 and 0.73 ± 0.14, respectively. By SHAP analysis, root mean square, mean attenuation, and kurtosis were the most predictive features in the texture features only model. Root mean square, cyst location, and mean attenuation were the most predictive features in the combined model.  
    Conclusion: Machine learning principles can be applied to PC texture features to create a mucinous phenotype classifier. Model performance did not improve with the combined model. However, specific radiomic, radiologic, and clinical features most predictive in our models can be identified using SHAP analysis.  
    Machine learning principles applied to CT radiomics to predict mucinous pancreatic cysts  
    Adam M. Awe et al.
    Abdominal Radiology (2022) 47:221–231
  • “Radiomics is the high throughput extraction of large sets of quantitative data from imaging studies that can be used to characterize healthy and pathological tissues to inform diagnosis and prognosis. Texture analysis, a subtype of radiomics, quantifies gray-level pixels and voxels in a frequency histogram and their spatial relationships to describe lesion heterogeneity within a 2-dimensional region of interest (ROI) or 3-dimensional volume of interest (VOI). Computed tomography (CT) texture analysis has demonstrated promise in diagnosing and risk-stratifying patients with PCs. Predictive ability of radiomics models can be enhanced by integrating clinical features in pancreas and non-pancreas tissues.”  
    Machine learning principles applied to CT radiomics to predict mucinous pancreatic cysts  
    Adam M. Awe et al.
    Abdominal Radiology (2022) 47:221–231
  • "Additional machine learning algorithms were applied to mucinous PCs to create a classifier that distinguishes cysts with HGD from cysts without HGD. Baseline models used for comparisons included minority, majority, random guesser, and stratified guesser models. XGBoost was also explored to evaluate their performance compared to baseline models. In developing the HGD classifier models, 5 and 9 weak learners were used in the decision trees for the texture features only and combined models, respectively. The XGBoost algorithms developed for the HGD classifier used a positive class weight scaling of 1.45 and 2.47, and maximum depth of 4 and 5 for the texture features only and combined models, respectively. The remainder of the decision tree parameters were left at their defaults. Accuracy, F1-score, and G-mean values were determined to compare performance of models.”
    Machine learning principles applied to CT radiomics to predict mucinous pancreatic cysts  
    Adam M. Awe et al.
    Abdominal Radiology (2022) 47:221–231
  • "In conclusion, our study demonstrates that machine learning principles can be applied to radiomics data of PCs to help detect mucinous phenotypes. While this information does not obviate the need for other diagnostic testing, it may help risk stratify patients with PCs. We also demonstrate that integration of radiologic and clinical features with texture feature radiomics data does not improve performance of our mucinous classifier. However, unique radiomic, radiologic, and clinical features were important in building our machine learning mucinous classifiers. These results highlight the potential of machine learning algorithms applied to high-throughput PC radiomics features in helping to detect mucinous cyst phenotype in patients and deserves further study to improve and validate such models.”
    Machine learning principles applied to CT radiomics to predict mucinous pancreatic cysts  
    Adam M. Awe et al.
    Abdominal Radiology (2022) 47:221–231
  • Background: At present, numerous challenges exist in the diagnosis of pancreatic SCNs and MCNs. After the emergence of artificial intelligence (AI), many radiomics research methods have been applied to the identification of pancreatic SCNs and MCNs. Purpose A deep neural network (DNN) model termed Multi-channel-Multiclassifier-Random Forest-ResNet (MMRF- ResNet) was constructed to provide an objective CT imaging basis for differential diagnosis between pancreatic serous cystic neoplasms (SCNs) and mucinous cystic neoplasms (MCNs).  
    Materials and methods: This study is a retrospective analysis of pancreatic unenhanced and enhanced CT images in 63 patients with pancreatic SCNs and 47 patients with MCNs (3 of which were mucinous cystadenocarcinoma) confirmed by pathology from December 2010 to August 2016. Different image segmented methods (single-channel manual outline ROI image and multi-channel image), feature extraction methods (wavelet, LBP, HOG, GLCM, Gabor, ResNet, and AlexNet) and classifiers (KNN, Softmax, Bayes, random forest classifier, and Majority Voting rule method) are used to classify the nature of the lesion in each CT image (SCNs/MCNs). Then, the comparisons of classification results were made based on sensitivity, specificity, precision, accuracy, F1 score, and area under the receiver operating characteristic curve (AUC), with pathological results serving as the gold standard.  
    Results: Multi-channel-ResNet (AUC 0.98) was superior to Manual-ResNet (AUC 0.91).CT image characteristics of lesions extracted by ResNet are more representative than wavelet, LBP, HOG, GLCM, Gabor, and AlexNet. Compared to the use of three classifiers alone and Majority Voting rule method, the use of the MMRF-ResNet model exhibits a better evaluation effect (AUC 0.96) for the classification of the pancreatic SCNs and MCNs.  
    Conclusion: The CT image classification model MMRF-ResNet is an effective method to distinguish between pancreatic SCNs and MCNs.  
    CT classification model of pancreatic serous cystic neoplasms and mucinous cystic neoplasms based on a deep neural network  
    Rong Yang et al.
    Abdominal Radiology (2022) 47:232–241 

  • CT classification model of pancreatic serous cystic neoplasms and mucinous cystic neoplasms based on a deep neural network  
    Rong Yang et al.
    Abdominal Radiology (2022) 47:232–241 
  • “In conclusion, in this study, Multi-channel CT images were obtained through preprocessing based on single-channel manual outline ROI images, and ResNet was used to extract CT image features of pancreatic SCNs and MCNs. The random forest classifier is used to integrate the classification probabilities of the KNN, Bayesian, and Softmax classifiers to determine the CT image properties of pancreatic SCNs and MCNs. Finally, a better classification result was obtained relative to the commonly used radiomics methods, suggesting that MMRF-ResNet is an ideal CT classification model for distinguishing between pancreatic SCNs and MCNs.”
    CT classification model of pancreatic serous cystic neoplasms and mucinous cystic neoplasms based on a deep neural network  
    Rong Yang et al.
    Abdominal Radiology (2022) 47:232–241 
  • Purpose: A deep neural network (DNN) model termed Multi-channel-Multiclassifier-Random Forest-ResNet (MMRF- ResNet) was constructed to provide an objective CT imaging basis for differential diagnosis between pancreatic serous cystic neoplasms (SCNs) and mucinous cystic neoplasms (MCNs).  
    Results: Multi-channel-ResNet (AUC 0.98) was superior to Manual-ResNet (AUC 0.91).CT image characteristics of lesions extracted by ResNet are more representative than wavelet, LBP, HOG, GLCM, Gabor, and AlexNet. Compared to the use of three classifiers alone and Majority Voting rule method, the use of the MMRF-ResNet model exhibits a better evaluation effect (AUC 0.96) for the classification of the pancreatic SCNs and MCNs.  
    Conclusion: The CT image classification model MMRF-ResNet is an effective method to distinguish between pancreatic SCNs and MCNs.
    CT classification model of pancreatic serous cystic neoplasms and mucinous cystic neoplasms based on a deep neural network  
    Rong Yang et al.
    Abdominal Radiology (2022) 47:232–241 
  • “we only analyzed the region of interest in images and did not analyze location information of the lesions (such as the head, body, and tail of the pancreas) and patient clinical information, such as gender, age, family history, and clinical symptoms, and the characteristics of the tumor have not been considered: size, grading, vascularization etc., for example are informations that can complete the clinical situation and they could be very useful notions.”
    CT classification model of pancreatic serous cystic neoplasms and mucinous cystic neoplasms based on a deep neural network  
    Rong Yang et al.
    Abdominal Radiology (2022) 47:232–241 
  •  “In conclusion, in this study, Multi-channel CT images were obtained through preprocessing based on single-channel manual outline ROI images, and ResNet was used to extract CT image features of pancreatic SCNs and MCNs. The random forest classifier is used to integrate the classification probabilities of the KNN, Bayesian, and Softmax classifiers to determine the CT image properties of pancreatic SCNs and MCNs. Finally, a better classification result was obtained relative to the commonly used radiomics methods, suggesting that MMRF-ResNet is an ideal CT classification model for distinguishing between pancreatic SCNs and MCNs.”
    CT classification model of pancreatic serous cystic neoplasms and mucinous cystic neoplasms based on a deep neural network  
    Rong Yang et al.
    Abdominal Radiology (2022) 47:232–241 
  • “In our study, the 2017 Fukuoka criteria performed slightly worse for selecting HRI for surgery compared to patients with sporadic cysts, missing 60% of cysts with invasive carcinoma, or IPMN with HGD, with a low sensitivity of 40%. Furthermore, the 2017 Fukuoka criteria might have resulted in unnecessary surgery of low-grade IPMN in our high risk population, with a modest specificity of 85%, which translates to ~15% of patients undergoing unnecessary or premature pancreatic surgery with its attendant morbidity and mortality. Similarly, the 2019 CAPS criteria missed 40% of resected IPMNs harboring advanced neoplasia while also recommending surgery for 15% of HRI that did not need it.”
    Guidelines on management of pancreatic cysts detected in high-risk individuals: An evaluation of the 2017 Fukuoka guidelines and the 2020 International Cancer of the Pancreas Screening (CAPS) consortium statements  
    Mohamad Dbouk , Olaya I. Brewer Gutierrez , Anne Marie Lennon , Miguel Chuidian , Eun Ji Shin , Ihab R. Kamel , Elliot K. Fishman , Jin He , Richard A. Burkhart , Christopher L. Wolfgang , Ralph H. Hruban , Michael G. Goggins, Marcia Irene Canto  
    Pancreatology 21 (2021) 613-621 
  • “One such approach is the detection of biomarkers in secretin- stimulated pancreatic juice at the time of EUS surveillance. The mutation profile and DNA concentration of pancreatic juice have been shown to be useful in the detection of high-grade PanIN le- sions and early PDAC in the CAPS cohort. Importantly, Yu et al. described the detection of low-abundance SMAD4/TP53 mu- tations from the cancer in the juice of patients under surveillance more than one year prior to the diagnosis of pancreatic mass on imaging. Overall, the analysis of pancreatic juice SMAD4/TP53 mutations could distinguish patients with PDAC or HGD from controls with a sensitivity and specificity of 72.2% and 89.4%, respectively.”
    Guidelines on management of pancreatic cysts detected in high-risk individuals: An evaluation of the 2017 Fukuoka guidelines and the 2020 International Cancer of the Pancreas Screening (CAPS) consortium statements  
    Mohamad Dbouk , Olaya I. Brewer Gutierrez , Anne Marie Lennon , Miguel Chuidian , Eun Ji Shin , Ihab R. Kamel , Elliot K. Fishman , Jin He , Richard A. Burkhart , Christopher L. Wolfgang , Ralph H. Hruban , Michael G. Goggins, Marcia Irene Canto  
    Pancreatology 21 (2021) 613-621 
  • “Finally, artificial intelligence and deep learning technologies applied to multi-detector pancreatic protocol CT may improve the early detection of pancreatic cancer or its precursor lesions.”
    Guidelines on management of pancreatic cysts detected in high-risk individuals: An evaluation of the 2017 Fukuoka guidelines and the 2020 International Cancer of the Pancreas Screening (CAPS) consortium statements  
    Mohamad Dbouk , Olaya I. Brewer Gutierrez , Anne Marie Lennon , Miguel Chuidian , Eun Ji Shin , Ihab R. Kamel , Elliot K. Fishman , Jin He , Richard A. Burkhart , Christopher L. Wolfgang , Ralph H. Hruban , Michael G. Goggins, Marcia Irene Canto  
    Pancreatology 21 (2021) 613-621 
  • “In conclusion, we report that the performance characteristics of the 2019 CAPS and 2017 Fukuoka ICG criteria for managing screen- detected pancreatic cysts are modestly specific but not sufficiently sensitive for selecting HRI for surgical treatment. New approaches, including multimodality algorithms that consider molecular cyst fluid analysis, clinical and genetic patient characteristics, and radiological pancreatic features, are needed to guide the surgical management of cystic lesions in individuals at high risk for pancreatic cancer.”
    Guidelines on management of pancreatic cysts detected in high-risk individuals: An evaluation of the 2017 Fukuoka guidelines and the 2020 International Cancer of the Pancreas Screening (CAPS) consortium statements  
    Mohamad Dbouk , Olaya I. Brewer Gutierrez , Anne Marie Lennon , Miguel Chuidian , Eun Ji Shin , Ihab R. Kamel , Elliot K. Fishman , Jin He , Richard A. Burkhart , Christopher L. Wolfgang , Ralph H. Hruban , Michael G. Goggins, Marcia Irene Canto  
    Pancreatology 21 (2021) 613-621 

  • Guidelines on management of pancreatic cysts detected in high-risk individuals: An evaluation of the 2017 Fukuoka guidelines and the 2020 International Cancer of the Pancreas Screening (CAPS) consortium statements  
    Mohamad Dbouk , Olaya I. Brewer Gutierrez , Anne Marie Lennon , Miguel Chuidian , Eun Ji Shin , Ihab R. Kamel , Elliot K. Fishman , Jin He , Richard A. Burkhart , Christopher L. Wolfgang , Ralph H. Hruban , Michael G. Goggins, Marcia Irene Canto  
    Pancreatology 21 (2021) 613-621 
  • ”Current PDAC surveillance is centered on imaging tests that detect pancreatic abnormalities, but as illustrated above, even individuals undergoing routine surveillance are sometimes diagnosed with advanced interval cancers.37 This fact suggests that improving the ability of imaging tests to identify precursor lesions could be useful in improving early detection. To that end, radiomics is an emerging field that holds promise. Traditionally, interpretation of radiographic images has been performed qualitatively by trained clinicians. The concept behind radiomics is to incorporate quantitative assessment of images by artificial intelligence, which allows for the identification of patterns not detectable visibly.”
    Pancreatic Cancer Surveillance and Novel Strategies for Screening  
    Beth Dudley, Randall E. Brand
    Gastrointest Endoscopy Clin N Am 32 (2022) 13–25 
  • Objective: To demonstrate enabling multi-institutional training without centralizing or sharing the underlying physical data via federated learning (FL). Materials and
    Methods: Deep learning models were trained at each participating institution using local clinical data, and an additional model was trained using FL across all of the institutions.  
    Results: We found that the FL model exhibited superior performance and generalizability to the models trained at single institutions, with an overall performance level that was significantly better than that of any of the institutional models alone when evaluated on held-out test sets from each institution and an outside challenge data- set.  
    Discussion: The power of FL was successfully demonstrated across 3 academic institutions while avoiding the privacy risk associated with the transfer and pooling of patient data.
    Conclusion: Federated learning is an effective methodology that merits further study to enable accelerated development of models across institutions, enabling greater generalizability in clinical use.  
    Federated learning improves site performance in multicenter deep learning without data sharing  
    Karthik V. Sarma et al.
    J Am Med Inform Assoc. 2021 Jun 12;28(6):1259-1264
  • “The power of federated learning was successfully demonstrated across 3 academic institutions using real clinical prostate imaging data. The federated model demonstrated improved performance across both held-out test sets from each institution and an external test set, validating the FL paradigm. This methodology could be ap- plied to a wide variety of DL applications in medical image analysis and merits further study to enable accelerated development of DL models across institutions, enabling greater generalizability in clinical use.”
    Federated learning improves site performance in multicenter deep learning without data sharing  
    Karthik V. Sarma et al.
    J Am Med Inform Assoc. 2021 Jun 12;28(6):1259-1264
  • “Diagnostic AI has not realized its potential to improve diagnostic performance because it has not focused on supporting the diagnostic journey. Knowing the direction of the pathway in a complex environment is important, but the essential decision is determining the next step. A shift to wayfinding AI could help achieve the synergy of human intelligence and AI to achieve diagnostic excellence.”
    Next-Generation Artificial Intelligence for Diagnosis  From Predicting Diagnostic Labels to “Wayfinding”
    Julia Adler-Milstein et al.  
    JAMA( Published online)December 9, 2021 
  • “In this paper, we consider a partially supervised setting, where cheap image-level annotations are provided for all the train- ing data, and the costly per-voxel annotations are only available for a subset of them. We propose an Inductive Attention Guidance Network (IAG-Net) to jointly learn a global image-level classifier for normal/PDAC classification and a local voxel-level classifier for semi-supervised PDAC segmentation. We instantiate both the global and the local classifiers by multiple instance learning (MIL), where the attention guidance, indicating roughly where the PDAC regions are, is the key to bridging them: For global MIL based normal/PDAC classification, attention serves as a weight for each instance (voxel) during MIL pooling, which eliminates the distraction from the background; For local MIL based semi-supervised PDAC segmentation, the attention guidance is inductive, which not only provides bag-level pseudo-labels to training data without per-voxel annotations for MIL training, but also acts as a proxy of an instance-level classifier.”
    Learning Inductive Attention Guidance for Partially Supervised Pancreatic Ductal Adenocarcinoma Prediction  
    Yan Wang , Peng Tang, Yuyin Zhou , Wei Shen, Elliot K. Fishman , Alan L. Yuille,  
    IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 40, NO. 10, OCTOBER 2021 

  • Learning Inductive Attention Guidance for Partially Supervised Pancreatic Ductal Adenocarcinoma Prediction  
    Yan Wang , Peng Tang, Yuyin Zhou , Wei Shen, Elliot K. Fishman , Alan L. Yuille,  
    IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 40, NO. 10, OCTOBER 2021 

  • Learning Inductive Attention Guidance for Partially Supervised Pancreatic Ductal Adenocarcinoma Prediction  
    Yan Wang , Peng Tang, Yuyin Zhou , Wei Shen, Elliot K. Fishman , Alan L. Yuille,  
    IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 40, NO. 10, OCTOBER 2021 
  • “This paper addresses the problem of PDAC prediction i.e., normal/PDAC classification and PDAC segmentation under the partially supervised setting. We present an Inductive Attention Guidance (IAG) strategy for learning a global image-level clas- sifier for normal/PDAC segmentation and a local instance-level classifier for semi-supervised PDAC segmentation, which enjoys the advantages of bridging the MIL-based global and local classifiers. We showed empirically on the JHMI dataset the superiority of the proposed IAG-Net for PDAC predic- tion, which is helpful to computer-assisted clinical diagnoses. Additionally, we verified the generality of IAG-Net on the pancreas tumor segmentation dataset in MSD challenge.”  
    Learning Inductive Attention Guidance for Partially Supervised Pancreatic Ductal Adenocarcinoma Prediction  
    Yan Wang , Peng Tang, Yuyin Zhou , Wei Shen, Elliot K. Fishman , Alan L. Yuille,  
    IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 40, NO. 10, OCTOBER 2021
  • “This study aimed to investigate the diagnostic ability of carcinoembryonic antigen (CEA), cytology, and artificial intelligence (AI) by deep learning using cyst fluid in differentiating malignant from benign cystic lesions. We retrospectively reviewed 85 patients who underwent pancreatic cyst fluid analysis of surgical specimens or endoscopic ultrasound-guided fine-needle aspiration specimens. AI using deep learning was used to construct a diagnostic algorithm. CEA, carbohydrate antigen 19-9, carbohydrate antigen 125, amylase in the cyst fluid, sex, cyst location, connection of the pancreatic duct and cyst, type of cyst, and cytology were keyed into the AI algorithm, and the malignant predictive value of the output was calculated. Area under receiver-operating characteristics curves for the diagnostic ability of malignant cystic lesions were 0.719 (CEA), 0.739 (cytology), and 0.966 (AI). In the diagnostic ability of malignant cystic lesions, sensitivity, specificity, and accuracy of AI were 95.7%, 91.9%, and 92.9%, respectively. AI sensitivity was higher than that of CEA (60.9%, p = 0.021) and cytology (47.8%, p = 0.001). AI accuracy was also higher than CEA (71.8%, p < 0.001) and cytology (85.9%, p = 0.210). AI may improve the diagnostic ability in differentiating malignant from benign pancreatic cystic lesions.”
    Diagnostic ability of artificial intelligence using deep learning analysis of cyst fluid in differentiating malignant from benign pancreatic cystic lesions  
    Yusuke Kurita et al.
    Scientific Reports | (2019) 9:6893 
  • "Although cytology had excellent specificity, it has a limited role because of its lack of sensitivity in previous studies30–32. In the present study, the sensitivity of cytology in differentiating malignant from benign cystic lesions was 47.8%. Thus, we constructed AI using deep learning algorithm for differentiating malignant from benign pancreatic cystic lesions based on the analysis of pancreatic cyst fluid and clinical data.”  
    Diagnostic ability of artificial intelligence using deep learning analysis of cyst fluid in differentiating malignant from benign pancreatic cystic lesions  
    Yusuke Kurita et al.
    Scientific Reports | (2019) 9:6893 
  • "In this study, AI using deep learning analyzed pancreatic cyst fluid and clinical data. By using this deep learning method, AI learns the characteristics of malignant cystic lesions by combining cyst fluid analysis and clinical data, and AI can possibly exclude the bias generated by human judgment. Although it is difficult for clinicians to diagnose malignant pancreatic cystic lesions by cyst fluid analysis and clinical data, AI using deep learning achieved adequate diagnostic ability in differentiating malignant from benign cystic lesions compared to cyst fluid analysis such as CEA and cytology. AI and CEA were also significant factor in the multivariate analysis of malignant cystic lesion. Specifically, although it is generally a problem that cytology diagnosis has low sensitivity, AI using deep learning achieved high sensitivity (95.7%).”
    Diagnostic ability of artificial intelligence using deep learning analysis of cyst fluid in differentiating malignant from benign pancreatic cystic lesions  
    Yusuke Kurita et al.
    Scientific Reports | (2019) 9:6893 
  • “Pancreatic cysts are common and often pose a management dilemma, because some cysts are precancerous, whereas others have little risk of developing into invasive cancers. We used supervised machine learning techniques to develop a comprehensive test, CompCyst, to guide the management of patients with pancreatic cysts. The test is based on selected clinical features, imaging characteristics, and cyst fluid genetic and biochemical markers. Using data from 436 patients with pancreatic cysts, we trained CompCyst to classify patients as those who required surgery, those who should be routinely monitored, and those who did not require further surveillance. We then tested CompCyst in an independent cohort of 426 patients, with histopathology used as the gold standard. We found that clinical management informed by the CompCyst test was more accurate than the management dictated by conventional clinical and imaging criteria alone. Application of the CompCyst test would have spared surgery in more than half of the patients who underwent unnecessary resection of their cysts. CompCyst therefore has the potential to reduce the patient morbidity and economic costs associated with current standard-of-care pancreatic cyst management practices.”
    A multimodality test to guide the management of patients with a pancreatic cyst.  
    Springer S, Masica DL, Dal Molin M, et al.  
    Sci Transl Med. 2019;11(501):eaav4772. doi:10.1126/scitranslmed.aav4772
  • “We found that clinical management informed by the CompCyst test was more accurate than the management dictated by conventional clinical and imaging criteria alone. Application of the CompCyst test would have spared surgery in more than half of the patients who underwent unnecessary resection of their cysts. CompCyst therefore has the potential to reduce the patient morbidity and economic costs associated with current standard-of-care pancreatic cyst management practices.”
    A multimodality test to guide the management of patients with a pancreatic cyst.  
    Springer S, Masica DL, Dal Molin M, et al.  
    Sci Transl Med. 2019;11(501):eaav4772. doi:10.1126/scitranslmed.aav4772
  • “Two dilemmas make pancreatic cyst clinical management challenging. First, it is difficult to differentiate IPMNs and MCNs, collectively termed “mucin-producing cysts,” from cysts that have no malignant potential and do not require any follow-up. Second, it can be difficult to differentiate patients with mucin-producing cysts that harbor early invasive cancer or high-grade dysplasia from patients with less advanced mucin-producing cysts. Surgery is recommended for patients with advanced cysts, whereas intermittent surveillance with imaging, rather than surgery, is considered appropriate for patients with less advanced cysts. Currently available clinical tools, however, are imperfect at assigning the most appropriate management strategies for patients with cysts. This is highlighted by the fact that 25% of cyst patients who undergo surgical resection have a pancreatic cyst with no malignant potential, and up to 78% of mucin-producing cysts referred for surgical resection are ultimately found not to be advanced, that is, they do not harbor high-grade dysplasia or cancer.”
    A multimodality test to guide the management of patients with a pancreatic cyst.  
    Springer S, Masica DL, Dal Molin M, et al.  
    Sci Transl Med. 2019;11(501):eaav4772. doi:10.1126/scitranslmed.aav4772

  • A multimodality test to guide the management of patients with a pancreatic cyst.  
    Springer S, Masica DL, Dal Molin M, et al.  
    Sci Transl Med. 2019;11(501):eaav4772. doi:10.1126/scitranslmed.aav4772

  • A multimodality test to guide the management of patients with a pancreatic cyst.  
    Springer S, Masica DL, Dal Molin M, et al.  
    Sci Transl Med. 2019;11(501):eaav4772. doi:10.1126/scitranslmed.aav4772

  • A multimodality test to guide the management of patients with a pancreatic cyst.  
    Springer S, Masica DL, Dal Molin M, et al.  
    Sci Transl Med. 2019;11(501):eaav4772. doi:10.1126/scitranslmed.aav4772
  •  “In conclusion, the use of a comprehensive test that evaluates clinical, imaging, and molecular features is imperfect but appears to offer substantial improvements over standard-of-care management of patients with pancreatic cysts. CompCyst does not replace conventional clinical tools. Instead, it contributes additional information, allowing clinicians to make more informed decisions. How and when tests like CompCyst can be implemented in routine clinical settings remains to be determined, but our results represent the next stage of research required for such implementation. An important next test of the markers presented here could be their validation in a follow-up, prospective study.”
    A multimodality test to guide the management of patients with a pancreatic cyst.  
    Springer S, Masica DL, Dal Molin M, et al.  
    Sci Transl Med. 2019;11(501):eaav4772. doi:10.1126/scitranslmed.aav4772
  • "Pancreatic cystic lesions, particularly IPMN, are the precursors of pancreatic cancer. Kuwahara et al. successfully established an AI-aided EUS using deep learning to distinguish malignant IPMNs from benign ones. The AI-aided EUS could diagnose malignant probability with a high sensitivity of 95.7% and a high accuracy of 94.0%, which was much greater than that of experts’ diagnoses (56.0%). AI-aided diagnosis is under development not only for IPMNs but also for other cystic lesions of the pancreas, such as serous cystic neoplasms, mucinous cystic neoplasms, solid pseudopapillary neoplasms, and cystic pancreatic neuroendocrine neoplasms.”
    A New Dawn for the Use of Artificial Intelligence in Gastroenterology, Hepatology and Pancreatology
    Akihiko Oka  , Norihisa Ishimura and Shunji Ishihara  
    Diagnostics 2021, 11, 1719. https://doi.org/10.3390/diagnostics11091719 
  • “Pancreatic schwannoma is a slowly growing, encapsulated, benign neoplasm that typically arises in the peripheral epineurium of either the sympathetic or parasympathetic autonomic fibers or branches of vagus nerve that extend to the pancreas. Pancreatic schwannomas most frequently involve the pancreas head (40%), followed by body (21%), neck (6%), tail (15%), and uncinate process (13%), respectively.”
    Abdominal schwannomas: review of imaging findings and pathology.
    Lee NJ, Hruban RH, Fishman EK.Abdom Radiol (NY).
    2017 Jul;42(7):1864-1870
  • "The features of pancreatic schwannomas on CT scan include low-density and/or cystic degenerative areas. MR imaging usually shows hypointensity on T1-weighted images and hyperintensity on T2-weighted images but like the CT features, these findings are nonspecific. Two-thirds of pancreatic schwannomas undergo degenerative changes such as cyst formation, necrosis, calcification, and hemorrhage, and these changes can mimic pancreatic cystic tumors.”
    Abdominal schwannomas: review of imaging findings and pathology.
    Lee NJ, Hruban RH, Fishman EK.Abdom Radiol (NY).
    2017 Jul;42(7):1864-1870
  • "In conclusion, 53.4% of patients diagnosed with clinical stage I PDAC demonstrated focal pancreatic abnormalities on pre-diagnostic CT examinations obtained at least one year before the diagnosis of PDAC. The most common focal abnormality on pre-diagnostic CT in patients who developed PDAC was focal parenchymal atrophy, followed by focal faint parenchymal enhancement and focal MPD change. Among these three findings, focal MPD change exhibited the shortest duration between its new development and the subsequent diagnosis of PDAC, while focal atrophy and faint enhancement exhibited more prolonged duration. These observations could facilitate earlier diagnosis of PDAC and thus improve management and prognosis.”
    CT Abnormalities of the Pancreas Associated With the Subsequent Diagnosis of Clinical Stage I Pancreatic Ductal Adenocarcinoma More Than One Year Later: A Case-Control Study  
    Fumihito Toshima et al.
    AJR 2021(in press) https://doi.org/10.2214/AJR.21.26014 
  • “Rigiroli et al provide an important advance in the struggle to select appropriate surgical candidates with pancreatic ductal adenocarcinoma based on preoperative CT imaging. With the increased use of neoadjuvant therapy, this article is particularly relevant given the known challenges in assessing vascular involvement after chemotherapy. However, the reality of neoadjuvant therapy is that the treatment landscape is evolving rapidly, with new drug and external beam radiation trials maturing every year. A persistent challenge in bringing radiomics to clinical practice in patients with cancer is the generalizability of predictive models that are derived from a subset of treatment regimens that may no longer be relevant over time.”
    Radiomics for CT Assessment of Vascular Contact in Pancreatic Adenocarcinoma  
    Richard K.G.Do, Avinash Kambadakone
    Radiology 2021; 00:1–2 • https://doi.org/10.1148/radiol.2021211635
  • Background: Current imaging methods for prediction of complete margin resection (R0) in patients with pancreatic ductal adeno- carcinoma (PDAC) are not reliable.  
    Purpose: To investigate whether tumor-related and perivascular CT radiomic features improve preoperative assessment of arterial involvement in patients with surgically proven PDAC.  
    Conclusion: A model based on tumor-related and perivascular CT radiomic features improved the detection of superior mesenteric artery involvement in patients with pancreatic ductal adenocarcinoma.  
    CT Radiomic Features of Superior Mesenteric Artery Involvement in Pancreatic Ductal Adenocarcinoma: A Pilot Study  
    Francesca Rigiroli et al.
    Radiology 2021; 000:1–13 • https://doi.org/10.1148/radiol.2021210699 
  • • In a retrospective study of 194 patients with pancreatic ductal adenocarcinoma, CT radiomic features demonstrated sensitivity of 62% (33 of 53 patients) and specificity of 77% (108 of 141 patients) in the detection of superior mesenteric artery involvement in patients undergoing surgery for pancreatic ductal adenocarcinoma.  
    • The radiomic model results outperformed the assessment made by expert radiologists in consensus during a multidisciplinary meeting, yielding areas under the curve of 0.71 and 0.54, respectively (P , .001).  
    CT Radiomic Features of Superior Mesenteric Artery Involvement in Pancreatic Ductal Adenocarcinoma: A Pilot Study  
    Francesca Rigiroli et al.
    Radiology 2021; 000:1–13 • https://doi.org/10.1148/radiol.2021210699 
  • “In conclusion, our results suggest that the analysis of tu- mor-related and perivascular radiomic features improves pre- operative assessment of tumor involvement of the superior mesenteric artery in patients with pancreatic ductal adenocar- cinoma, a highly challenging task for even experienced multi- disciplinary teams, particularly after neoadjuvant therapy. To ensure our model is valid and unbiased, it should be validated in a separate independent data set. Future work may also in- corporate more sophisticated modeling techniques, including unsupervised machine learning frameworks and deep learning algorithms that fuse radiomics data with other types of clinical data.”
    CT Radiomic Features of Superior Mesenteric Artery Involvement in Pancreatic Ductal Adenocarcinoma: A Pilot Study  
    Francesca Rigiroli et al.
    Radiology 2021; 000:1–13 • https://doi.org/10.1148/radiol.2021210699 
  • “In our study, the radiomic model showed higher negative predictive value than the multidisciplinary assessment in ruling out SMA tumoral involvement defined by a clearance of 1 mm. Our results and previous studies have shown that predicting margin status using only standard CT criteria is challenging. Recent investigations have emphasized the need for optimal identification of patients with high likelihood of margin- negative resection, such as with a tumor more than 1 mm from the margin, because it yields a better prognosis compared with patients with positive surgical margin (tumor ≤1 mm to the margin or direct involvement). Despite being limited to assessment of the SMA margin, the application of our radiomic model in a clinical setting could help to guide radiologists in predicting margin status.”
    CT Radiomic Features of Superior Mesenteric Artery Involvement in Pancreatic Ductal Adenocarcinoma: A Pilot Study  
    Francesca Rigiroli et al.
    Radiology 2021; 000:1–13 • https://doi.org/10.1148/radiol.2021210699 
  • OBJECTIVE. Pancreatic ductal adenocarcinoma (PDAC) is often a lethal malignancy with limited preoperative predictors of long-term survival. The purpose of this study was to evaluate the prognostic utility of preoperative CT radiomics features in predict- ing postoperative survival of patients with PDAC.  
    RESULTS. The mean age of patients with PDAC was 67 ± 11 (SD) years. The mean tu- mor size was 3.31 ± 2.55 cm. The 10 most relevant radiomics features showed 82.2% ac- curacy in the classification of high-risk versus low-risk groups. The C-index of survival prediction with clinical parameters alone was 0.6785. The addition of CT radiomics features improved the C-index to 0.7414.  
    CONCLUSION. Addition of CT radiomics features to standard clinical factors im- proves survival prediction in patients with PDAC.  
    CT Radiomics–Based Preoperative Survival Prediction in Patients With Pancreatic Ductal Adenocarcinoma  
    Seyoun P, Kawamoto S, Fishman EK, Chu LC et al.
    AJR:217, November 2021 (in press) 
  • “Survival time after the surgical resection was used to stratify patients into a low- risk group (survival time > 3 years) and a high-risk group (survival time < 1 year). The 3D volume of the whole pancreatic tumor and background pancreas were manually seg- mented. A total of 478 radiomics features were extracted from tumors and 11 extra features were computed from pancreas boundaries. The 10 most relevant features were selected by feature reduction. Survival analysis was performed on the basis of clinical parameters both with and without the addition of the selected features. Survival status and time were estimated by a random survival forest algorithm. Concordance index (C-index) was used to evaluate performance of the survival prediction model.”
    CT Radiomics–Based Preoperative Survival Prediction in Patients With Pancreatic Ductal Adenocarcinoma  
    Seyoun P, Kawamoto S, Fishman EK, Chu LC et al.
    AJR:217, November 2021 (in press) 
  • "Tumors are spatially heterogeneous structures that can be characterized at a macro scale, and normal parenchyma can also be affected by the growth of the tumor. Texture analysis using medical images, especially radiomics approaches, is an established tech- nique that describes spatial variations in pixel intensities in images for quantitative assessment. Whereas radiologists may qualitatively describe PDAC enhancement patterns as, for example, homogeneously isoattenuating or heterogeneously hypoattenuating, tex- ture analysis can capture more subtle underlying differences that may reflect important pathologic differences and thereby help predict patient survival.”
    CT Radiomics–Based Preoperative Survival Prediction in Patients With Pancreatic Ductal Adenocarcinoma  
    Seyoun P, Kawamoto S, Fishman EK, Chu LC et al.
    AJR:217, November 2021 (in press) 
  •  “The entire 3D volume of the pancreas was segmented based on thin-slice venous phase images. The 3D volume of the whole tumor and background pancreas was manually segmented by four trained researchers using a commercial annotation software (Velocity, Varian Medical Systems). The boundaries were verified by three abdominal radiologists with 5–30 years of experience. Each case was randomly assigned to one of the researchers and a radiologist. The researcher and radiologist had face-to-face sessions to review each case. Any disagreement or errors identified during this review were corrected.”
    CT Radiomics–Based Preoperative Survival Prediction in Patients With Pancreatic Ductal Adenocarcinoma  
    Seyoun P, Kawamoto S, Fishman EK, Chu LC et al.
    AJR:217, November 2021 (in press) 
  • "Based on the selected radiomics features, a random survival forest was applied for survival prediction in a multivariate dataset with missing variables. Each decision node was divided until three unique deaths (d = 3) remained in the leaf node. Ten thousand trees were built by the training set using the AUC for the split of internal nodes. Each end node stored the survival sta- tus (dead or alive), survival time, and a Cox proportional hazard function of the assigned cases. The survival time and survival status predictions in the validation cohort were determined by majority voting based on the trained trees.”
    CT Radiomics–Based Preoperative Survival Prediction in Patients With Pancreatic Ductal Adenocarcinoma  
    Seyoun P, Kawamoto S, Fishman EK, Chu LC et al.
    AJR:217, November 2021 (in press)  

  • CT Radiomics–Based Preoperative Survival Prediction in Patients With Pancreatic Ductal Adenocarcinoma  
    Seyoun P, Kawamoto S, Fishman EK, Chu LC et al.
    AJR:217, November 2021 (in press) 
  • "The 10 most relevant radiomics features were selected to distinguish the high- and low-risk groups and are listed in Table 4. To determine the classification power of the selected features, binary classification was performed using a random forest. Among 90 patients, 45 (50.0%, 22 low-risk and 23 high-risk) randomly select- ed patients were included in the training set, and the remaining 45 (50.0%, 23 low-risk and 22 high-risk) were included in the validation set. The overall accuracy of classification of patients into high- and low-risk groups based on selected image features was 82.2%. The high-risk group showed a higher classification performance, with 86.4% accuracy, compared with the low-risk group, with 78.3% accuracy.”
    CT Radiomics–Based Preoperative Survival Prediction in Patients With Pancreatic Ductal Adenocarcinoma  
    Seyoun P, Kawamoto S, Fishman EK, Chu LC et al.
    AJR:217, November 2021 (in press) 
  •  "We found that radiomics features extracted from tumors and from the nonneoplastic pancreas can be used to improve survival prediction models of patients who underwent surgery for PDAC. This algorithm could be combined with other pathologic and genetic biomarkers.”
    CT Radiomics–Based Preoperative Survival Prediction in Patients With Pancreatic Ductal Adenocarcinoma  
    Seyoun P, Kawamoto S, Fishman EK, Chu LC et al.
    AJR:217, November 2021 (in press)  

  • CT Radiomics–Based Preoperative Survival Prediction in Patients With Pancreatic Ductal Adenocarcinoma  
    Seyoun P, Kawamoto S, Fishman EK, Chu LC et al.
    AJR:217, November 2021 (in press) 
  • “Pancreatic ductal adenocarcinoma (PDAC) is a leading cause of mortality among all cancers. It ranked fifth among all cancers in terms of mortality and its overall 5-year survival rate was just 6% in Korea for 2015. Surgical resection is essential for its cure but only a small proportion of its cases are found at an early stage enough for the procedure. Moreover, its recurrence rate after surgery is estimated to be 50%–60%, while its 5-year survival rate after surgery is reported to be just 20%–30% . The mean disease-free period in imaging studies is 267 ± 158 d with negative surgical margins, but 72 ± 47 d with positive margins. Therefore the survival of patients with PDAC is closely related to recurrence, and recurrence after surgery is one of the typical characteristics of PDAC .”
    Usefulness of artificial intelligence for predicting recurrence following surgery for pancreatic cancer: Retrospective cohort study  
    Kwang-Sig Lee et al.
    International Journal of Surgery 93 (2021) 106050 
  • "It is very important to prevent the recurrence of pancreatic cancer after surgery and there has been strong endeavor to identify major predictors of its disease-free survival after surgery. However, the results of existing literature were inconsistent and predictors in these studies were unmodifiable in general. Predictive nomograms were developed to combine and visualize the findings of traditional statistical models such as logistic regression and the Cox model regarding the recurrence of pancreatic cancer after surgery. But the predictive nomograms still require unrealistic assumptions of the traditional statistical models, i.e., ceteris paribus, “all the other variables staying constant”. In this context, this study used the random forest and multi-center registry data to analyze the recurrence of pancreatic cancer after surgery and its major determinants.”
    Usefulness of artificial intelligence for predicting recurrence following surgery for pancreatic cancer: Retrospective cohort study  
    Kwang-Sig Lee et al.
    International Journal of Surgery 93 (2021) 106050 
  • "Secondly, it was beyond the scope of this study to combine deep learning and the Cox model for predicting the recurrence of pancreatic cancer after surgery. Deep learning can be defined as “a sub-group of the artificial neural network whose number of hidden layers is larger than five, e.g., ten”. The last three years have seen the emergence of new strands of research to combine the Cox model with different types of its deep-learning counterparts. The continued development and application of these cutting-edge approaches would break new ground and bring more profound clinical insights regarding the recurrence of pancreatic cancer after surgery and its major determinants.”
    Usefulness of artificial intelligence for predicting recurrence following surgery for pancreatic cancer: Retrospective cohort study  
    Kwang-Sig Lee et al.
    International Journal of Surgery 93 (2021) 106050 
  • “This study aimed to investigate the diagnostic ability of carcinoembryonic antigen (CEA), cytology, and artificial intelligence (AI) by deep learning using cyst fluid in differentiating malignant from benign cystic lesions. We retrospectively reviewed 85 patients who underwent pancreatic cyst fluid analysis of surgical specimens or endoscopic ultrasound-guided fine-needle aspiration specimens. AI using deep learning was used to construct a diagnostic algorithm. CEA, carbohydrate antigen 19-9, carbohydrate antigen 125, amylase in the cyst fluid, sex, cyst location, connection of the pancreatic duct and cyst, type of cyst, and cytology were keyed into the AI algorithm, and the malignant predictive value of the output was calculated. Area under receiver-operating characteristics curves for the diagnostic ability of malignant cystic lesions were 0.719 (CEA), 0.739 (cytology), and 0.966 (AI). In the diagnostic ability of malignant cystic lesions, sensitivity, specificity, and accuracy of AI were 95.7%, 91.9%, and 92.9%, respectively. AI sensitivity was higher than that of CEA (60.9%, p = 0.021) and cytology (47.8%, p = 0.001). AI accuracy was also higher than CEA (71.8%, p < 0.001) and cytology (85.9%, p = 0.210). AI may improve the diagnostic ability in differentiating malignant from benign pancreatic cystic lesions.”
    Diagnostic ability of artificial intelligence using deep learning analysis of cyst fluid in differentiating malignant from benign pancreatic cystic lesions  
    Yusuke Kurita et al.
    Scientific Reports | (2019) 9:6893 
  • "Although cytology had excellent specificity, it has a limited role because of its lack of sensitivity in previous studies30–32. In the present study, the sensitivity of cytology in differentiating malignant from benign cystic lesions was 47.8%. Thus, we constructed AI using deep learning algorithm for differentiating malignant from benign pancreatic cystic lesions based on the analysis of pancreatic cyst fluid and clinical data.”  
    Diagnostic ability of artificial intelligence using deep learning analysis of cyst fluid in differentiating malignant from benign pancreatic cystic lesions  
    Yusuke Kurita et al.
    Scientific Reports | (2019) 9:6893 
  • "In this study, AI using deep learning analyzed pancreatic cyst fluid and clinical data. By using this deep learning method, AI learns the characteristics of malignant cystic lesions by combining cyst fluid analysis and clinical data, and AI can possibly exclude the bias generated by human judgment. Although it is difficult for clinicians to diagnose malignant pancreatic cystic lesions by cyst fluid analysis and clinical data, AI using deep learning achieved adequate diagnostic ability in differentiating malignant from benign cystic lesions compared to cyst fluid analysis such as CEA and cytology. AI and CEA were also significant factor in the multivariate analysis of malignant cystic lesion. Specifically, although it is generally a problem that cytology diagnosis has low sensitivity, AI using deep learning achieved high sensitivity (95.7%).”
    Diagnostic ability of artificial intelligence using deep learning analysis of cyst fluid in differentiating malignant from benign pancreatic cystic lesions  
    Yusuke Kurita et al.
    Scientific Reports | (2019) 9:6893 
  • "Although only a few studies describing the use of radiomics in risk stratification of PCLs have been published, these studies have demonstrated that radiomics can be utilized to non-invasively discriminate between low-risk and high-risk PCLs before resection. This cost-effective approach would enable us to accurately recommend lifesaving surgery for individuals with malignant cysts and spare those with benign lesions the morbidity, mortality and high costs associated with pancreatic surgeries. Consequently, more studies are warranted to develop these imaging biomarkers which can be used to differentiate between benign and malignant PCLs.”
    Radiomics in stratification of pancreatic cystic lesions: Machine learning in action  
    Vipin Dalala et al.
    Cancer Letters,Volume 469,2020,Pages 228-237
  • "Chakraborty et al. utilized radiomics features extracted from pre- surgical CT images, as markers for assessment of malignancy risk of BD- IPMNs. Similar to the previous studies, they categorized their cohort of 103 patients into low-risk and high-risk IPMNs based on final pathological findings after cyst resection. They extracted four new radio- graphically inspired features (enhanced boundary fraction, enhanced inside fraction, filled largest connected component fraction and average weighted eccentricity), along with intensity and orientation-based texture features from the CT images.”
    Radiomics in stratification of pancreatic cystic lesions: Machine learning in action  
    Vipin Dalala et al.
    Cancer Letters,Volume 469,2020,Pages 228-237
  • "This has led to an increased interest in radiomics, a high-throughput extraction of comprehensible data from standard of care images. Radiomics can be used as a diagnostic and prognostic tool in personalized medicine. It utilizes quantitative image analysis to extract features in conjunction with machine learning and artificial intelligence (AI) methods like support vector machines, random forest, and convolutional neural network for feature se- lection and classification. Selected features can then serve as imaging biomarkers to predict high-risk PCLs. Radiomics studies conducted heretofore on PCLs have shown promising results.”
    Radiomics in stratification of pancreatic cystic lesions: Machine learning in action  
    Vipin Dalala et al.
    Cancer Letters,Volume 469,2020,Pages 228-237
  • "IPMNs and MCNs are the only radiographically identifiable precursors of pancreatic cancer. Consequently, accurate assessment of the malignant potential of these cystic lesions may allow early detection of resectable PCLs prior to oncogenesis. The latest guidelines propose a practical approach for their management and surveillance, yet the clinical management of these mucinous cystic lesions remains challenging. The variable risk of malignant transformation combined with elevated risks associated with pancreatic surgery have led to conflicting recommendations for the management of mucinous cystic lesions.”
    Radiomics in stratification of pancreatic cystic lesions: Machine learning in action  
    Vipin Dalala et al.
    Cancer Letters,Volume 469,2020,Pages 228-237
  • “The multivariable logistic regression model included sex, size, location, shape, cyst characteristic, and cystic wall thickening. The individualized prediction nomogram showed good discrimination in the training sample (AUC 0.89; 95% CI 0.83–0.95) and in the validation sample (AUC 0.81; 95% CI 0.70–0.94). If the threshold probability is between 0.03 and 0.9, and > 0.93 in the prediction model, using the nomogram to predict SCN and MCN is more beneficial than the treat-all- patients as SCN scheme or the treat-all-patients as MCN scheme. The prediction model showed better discrimination than the radiologists’ diagnosis (AUC = 0.68).”
    A nomogram for predicting pancreatic mucinous cystic neoplasm and serous cystic neoplasm  
    Chengwei Shao et al.
    Abdominal Radiology https://doi.org/10.1007/s00261-021-03038-3 
  • All tumors were evaluated for the following characteristics: (1) CT-reported tumor size (i.e., the maximum cross-sectional diameter of the tumor [13]); (2) tumor location: pancreatic head, body, or tail; (3) shape: round or lobulated (lobulation was defined as the presence of rounded contours that could not be described as the borders of the same circle [9]); (4) cyst characteristic: oligocystic or polycystic; (5) cystic wall: thin or thick (thin was defined as < 2 mm while thick was defined as ≥ 2 mm [9]); (6) calcification; (7) enhanced mural nodule; (8) parenchymal atrophy; (9) common bile duct cutoff and dila- tion (> 10 mm); (10) main pancreatic duct (MPD) cutoff and dilation (> 3 mm); (11) pancreatitis identified by stranding of the peripancreatic fat tissue, ill-defined parenchymal contours, and fluid collections in the peripancreatic region; (12) contour abnormality; and (13) number of lesions: 1 or ≥ 2.  
    A nomogram for predicting pancreatic mucinous cystic neoplasm and serous cystic neoplasm  
    Chengwei Shao et al.
    Abdominal Radiology https://doi.org/10.1007/s00261-021-03038-3 
  • “There were several limitations to this study. First, the number of patients was relatively small. Second, this was a single-center, retrospective analysis. In the future, we will expand the number of cases and perform a multi-center validation of the model. Third, the predicted model in this study only focused on SCN and MCN, and did not include other cystic lesions of the pancreas such as IPMN, pseudocyst, and retention cyst. Lastly, we only used CT characteristics to develop the model. We did not combine radiomics features, although artificial intelligence is becoming a hot topic. In the future, we will combine the CT characteristics and radiomics features to develop a more accurate model.”
    A nomogram for predicting pancreatic mucinous cystic neoplasm and serous cystic neoplasm  
    Chengwei Shao et al.
    Abdominal Radiology https://doi.org/10.1007/s00261-021-03038-3 
  • “Lastly, we only used CT characteristics to develop the model. We did not combine radiomics features, although artificial intelligence is becoming a hot topic. In the future, we will combine the CT characteristics and radiomics features to develop a more accurate model.”
    A nomogram for predicting pancreatic mucinous cystic neoplasm and serous cystic neoplasm  
    Chengwei Shao et al.
    Abdominal Radiology https://doi.org/10.1007/s00261-021-03038-3  
  • “Pancreatic cancer (pancreatic ductal adenocarcinoma [PDAC]) is associated with a dire prognosis and a 5-year survival rate of only 10%. This statistic is somewhat misleading given that 52% of the patients will develop metastatic disease, with a resulting 2.9%, 5-year relative survival rate. However, for those patients with localized cancer where the tumor is confined to the primary site, the 5-year relative survival rate is 39.4%. It is estimated that in 2020, there will be 57,600 new cases of PDAC  and an estimated 47,050 will die of this disease.”  
    Artificial Intelligence and Early Detection of Pancreatic Cancer: 2020 Summative Review  
    Barbara Kenner, PhD,* Suresh T. Chari, MD,† David Kelsen, MD, Fishman EK et al.
    Pancreas. 2021 Mar 1;50(3):251-279
  • "Pancreatic ductal adenocarcinoma has the poorest overall survival of all the major cancer types, with a 5-year relative  survival rate that just reached 10%. This is due in part to the latestage at presentation, so that 49.6% of cases of newly diagnosed PDAC present with distant metastases, 29.1% present with re- gional lymph node involvement, and only 10.8% have tumors that are localized solely within the pancreas.”
    Artificial Intelligence and Early Detection of Pancreatic Cancer: 2020 Summative Review  
    Barbara Kenner, PhD,* Suresh T. Chari, MD,† David Kelsen, MD, Fishman EK et al.
    Pancreas. 2021 Mar 1;50(3):251-279

  • Artificial Intelligence and Early Detection of Pancreatic Cancer: 2020 Summative Review  
    Barbara Kenner, PhD,* Suresh T. Chari, MD,† David Kelsen, MD, Fishman EK et al.
    Pancreas. 2021 Mar 1;50(3):251-279
  • "In this context, the big data field provides a conceptual framework for analysis across the full spectrum of disease that may better capture patient subcategories, in particular when considering longitudinal disease development in a lifelong perspective. Here, variation in “healthy” diagnosis-free routes toward disease and later differences in disease comorbidities are currently of high interest. Using health care sector, socioeconomic, and consumer data, the precision medicine field works increasingly toward such a disease spectrum-wide approach. Ideally, this involves data describing healthy individuals, many of whom will later become sick—to have long-range correlations that relate to outcomes available for analysis. This notion extends the traditional disease trajectory concept into healthy life-course periods potentially enabling stratification of patient cohorts by systematically observed differences present before the onset and diagnosis of disease.”
    Artificial Intelligence and Early Detection of Pancreatic Cancer: 2020 Summative Review  
    Barbara Kenner, PhD,* Suresh T. Chari, MD,† David Kelsen, MD, Fishman EK et al.
    Pancreas. 2021 Mar 1;50(3):251-279
  • "Ultimately, it is likely that AI will transform much of the practice of medicine. AI will be used to interpret radiographs, ultrasounds, CT, and MRI, either as an adjunct to the clinician's interpretation or as the standalone reading.88 Health care organizations will use AI systems to extract and analyze electronic health record (EHR) data to better allocate staff and other resources, identify patients at risk for acute decompensation, and prevent medication errors.148 Using sensors on commodity devices such as smartphones, wearables, smart speakers, laptops, and tablets, individuals will be able to share health data during their daily lives and help generate a longitudinal personal health record, with pertinent information incorporated into their EHR. By extracting information from the EHR and incorporating data during an encounter with a patient, clinicians can be provided with a differential diagnosis in real-time with probabilities included.”
    Artificial Intelligence and Early Detection of Pancreatic Cancer: 2020 Summative Review  
    Barbara Kenner, PhD,* Suresh T. Chari, MD,† David Kelsen, MD, Fishman EK et al.
    Pancreas. 2021 Mar 1;50(3):251-279
  • "Because of the “black box” quality of many deep learning algorithms, clinicians and patients may be hesitant to depend on AI-based solutions. This fear is not unfounded. For example, it was discovered that an algorithm evaluating data from images of skin lesions was more likely to classify the lesion as malignant if a ruler was included in the photograph.149 The reticence by clinicians to embrace AI-based medical devices may also be explained by the paucity of peer-reviewed prospective studies assessing the efficacy of these systems.Finally, regulatory assessment of the effectiveness and safety of AI-based products is different from that of traditional medical devices.Regulatory agencies are working to find the best processes for determining whether an AI medical device should be cleared for clinical use.”
    Artificial Intelligence and Early Detection of Pancreatic Cancer: 2020 Summative Review  
    Barbara Kenner, PhD,* Suresh T. Chari, MD,† David Kelsen, MD, Fishman EK et al.
    Pancreas. 2021 Mar 1;50(3):251-279
  • "The ability to reliably detect very early-stage PDAC in asymptomatic patients should result in a major improvement in survival. This hypothesis is based on the observation that the prognosis for PDAC is clearly related to the pathological stage of the tumor at the time of diagnosis. Using the SEER database, Ansari et al reported that 5-year survival for patients with lymph node–negative primary PDAC less than 1-cm cancers is ~60%; with primary tumors of 2 cm or larger even without lymph node metastasis, survival was less than 20%. However, less than 1% of patients are found with primary PDAC less than 1 centimeter in size. Pancreatic ductal adenocarcinoma is diagnosed in the large majority of even stage IA patients because of symptoms, not as a result of an early detection program. The hypothesis that the earlier the stage of a PDAC, the better the outcome, is in concert with data from many other solid tumors, including breast, non–small cell lung, colorectal, prostate, and gastric cancers.”
    Artificial Intelligence and Early Detection of Pancreatic Cancer: 2020 Summative Review  
    Barbara Kenner, PhD,* Suresh T. Chari, MD,† David Kelsen, MD, Fishman EK et al.
    Pancreas. 2021 Mar 1;50(3):251-279
  • "Project Felix is a Lustgarten Foundation initiative led by Elliott Fishman at Johns Hopkins University to develop deep learning tools that can detect pancreatic tumors when they are smaller and with greater reliability than human readers alone. This effort has involved meticulous manual segmentation of thousands of abdominal CT scans to serve as a training and testing cohort, which represents the largest effort in this domain in the world. In collaboration with the computer scientist Alan Yuille. Project Felix has produced at least 17 articles on techniques to automatically detect and characterize lesions within the pancreas (https://www.ctisus.com/responsive/deep-learning/felix.asp).”
    Artificial Intelligence and Early Detection of Pancreatic Cancer: 2020 Summative Review  
    Barbara Kenner, PhD,* Suresh T. Chari, MD,† David Kelsen, MD, Fishman EK et al.
    Pancreas. 2021 Mar 1;50(3):251-279
  • "Eugene Koay from The University of Texas MD Anderson Cancer Center (MDACC) has previously characterized subtypes of PDAC on CT scans, whereby conspicuous (high delta) PDAC tumors are more likely to have aggressive biology, a higher rate of common pathway mutations, and poorer clinical outcomes compared with inconspicuous (low delta) tumors.His group has recently completed an analysis, currently under review, that shows that high-delta tumors demonstrate higher growth rates and shorter initiation times than their low-delta counterparts in the prediagnostic period. Although not strictly an AI initiative, his work serves as a rich foundation for future AI initiatives in this space. Drs Koay and Anirban Maitra at the MDACC are leading the NCI-sponsored EDRN initiative to assemble a prediagnosis pancreatic cancer cohort that could facilitate AI research into screening and early detection.”
    Artificial Intelligence and Early Detection of Pancreatic Cancer: 2020 Summative Review  
    Barbara Kenner, PhD,* Suresh T. Chari, MD,† David Kelsen, MD, Fishman EK et al.
    Pancreas. 2021 Mar 1;50(3):251-279
  • “Pancreatic cancer remains a major health problem, and only less than 20% of patients have resectable disease at the time of initial diagnosis. Systemic chemotherapy is often used in the patients with borderline resectable, locally advanced unresectable disease and metastatic disease. CT is often used to assess for therapeutic response; however, conventional imaging including CT may not correctly reflect treatment response after chemotherapy.”
    Assessment of iodine uptake by pancreatic cancer following chemotherapy using dual-energy CT.  
    Kawamoto S, Fuld MK, Laheru D, Huang P, Fishman EK.  
    Abdom Radiol (NY). 2018;43(2):445-456. 
  • "Dual-energy (DE) CT can acquire datasets at two different photon spectra in a single CT acquisition, and permits separating materials and extract iodine by applying a material decomposition algorithm. Quantitative iodine mapping may have an added value over conventional CT imaging for monitoring the treatment effects in patients with pancreatic cancer and potentially serve as a unique biomarker for treatment response. In this pictorial essay, we will review the technique for iodine quantification of pancreatic cancer by DECT and discuss our observations of iodine quantification at baseline and after systemic chemotherapy with conventional cytotoxic agents.”
    Assessment of iodine uptake by pancreatic cancer following chemotherapy using dual-energy CT.  
    Kawamoto S, Fuld MK, Laheru D, Huang P, Fishman EK.  
    Abdom Radiol (NY). 2018;43(2):445-456. 
  • “The parameters obtained using tumor segmentation software included (1) RECIST diameter (mm), (2) tumor volume (mL), (3) mean CT number of tumor (HU) at simulated weighted-average 120-kVp images, (4) iodine uptake by tumor per volume of tissue (mg/mL), and (5) normalized tumor iodine uptake (tumor iodine uptake normalized to the reference value acquired using region of interest place in the abdominal aorta at the level of the pancreatic tumor, calculated by tumor iodine uptake [mg/dL]/abdominal aortic uptake [mg/dL]).”
  • “In conclusion, iodine uptake by pancreatic adenocarcinoma using DECT may add supplemental information for assessment of treatment response, although tumor iodine uptake by pancreatic adenocarcinoma is small, and it may be difficult to apply to each case. Normalized tumor iodine uptake might be more sensitive than iodine concentration to measure treatment response. More data are necessary to confirm these observations.”
    Assessment of iodine uptake by pancreatic cancer following chemotherapy using dual-energy CT.  
    Kawamoto S, Fuld MK, Laheru D, Huang P, Fishman EK.  
    Abdom Radiol (NY). 2018;43(2):445-456.
  • Purpose: Evaluate utility of dual energy CT iodine material density images to identify preoperatively nodal positivity in pancreatic cancer patients who underwent neoadjuvant therapy.
    Conclusion: The dual energy based minimum normalized iodine value of all nodes in the surgical field on preoperative studies has modest utility in differentiating N0 from N1/2, and generally outperformed conventional features for identifying nodal metastases.
    CT features predictive of nodal positivity at surgery in pancreatic cancer patients following neoadjuvant therapy in the setting of dual energy CT.  
    Le O, Javadi S, Bhosale PR et al.  
    Abdom Radiol (NY). 2021 Jan 20. doi: 10.1007/s00261-020-02917-5. Epub ahead of print. PMID: 33471129.
  • Background: The diagnostic performance of CT for pancreatic cancer is interpreter-dependent, and approximately 40% of tumours smaller than 2 cm evade detection. Convolutional neural networks (CNNs) have shown promise in image analysis, but the networks’ potential for pancreatic cancer detection and diagnosis is unclear. We aimed to investigate whether CNN could distinguish individuals with and without pancreatic cancer on CT, compared with radiologist interpretation. Interpretation CNN could accurately distinguish pancreatic cancer on CT, with acceptable generalisability to images of patients from various races and ethnicities. CNN could supplement radiologist interpretation.
    Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation
    Kao-Lang Liu et al.
    Lancet Digital Health 2020; 2: e303–13
  • “Findings Between Jan 1, 2006, and Dec 31, 2018, we obtained CT images. In local test set 1, CNN-based analysis had a sensitivity of 0·973, specificity of 1·000, and accuracy of 0·986 (area under the curve [AUC] 0·997 (95% CI 0·992–1·000). In local test set 2, CNN-based analysis had a sensitivity of 0·990, specificity of 0·989, and accuracy of 0·989 (AUC 0·999 [0·998–1·000]). In the US test set, CNN-based analysis had a sensitivity of 0·790, specificity of 0·976, and accuracy of 0·832 (AUC 0·920 [0·891–0·948)]. CNN-based analysis achieved higher sensitivity than radiologists did (0·983 vs 0·929, difference 0·054 [95% CI 0·011–0·098]; p=0·014) in the two local test sets combined. CNN missed three (1·7%) of 176 pancreatic cancers (1·1–1·2 cm). Radiologists missed 12 (7%) of 168 pancreatic cancers (1·0–3·3 cm), of which 11 (92%) were correctly classified using CNN. The sensitivity of CNN for tumours smaller than 2 cm was 92·1% in the local test sets and 63·1% in the US test set.”
    Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation
    Kao-Lang Liu et al.
    Lancet Digital Health 2020; 2: e303–13
  • Findings: CNN-based analysis achieved higher sensitivity than radiologists did (0·983 vs 0·929, difference 0·054 [95% CI 0·011–0·098]; p=0·014) in the two local test sets combined. CNN missed three (1·7%) of 176 pancreatic cancers (1·1–1·2 cm). Radiologists missed 12 (7%) of 168 pancreatic cancers (1·0–3·3 cm), of which 11 (92%) were correctly classified using CNN. The sensitivity of CNN for tumours smaller than 2 cm was 92·1% in the local test sets and 63·1% in the US test set.
    Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation
    Kao-Lang Liu et al.
    Lancet Digital Health 2020; 2: e303–13
  • “CNN can accurately differentiate pancreatic cancer from non-cancerous pancreas, and with improvements might accommodate variations in patient race and ethnicity and imaging parameters that are inevitable in real-world clinical practice. CNN holds promise for developing computer-aided detection and diagnosis tools for pancreatic cancer to supplement radiologist interpretation.”
    Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation
    Kao-Lang Liu et al.
    Lancet Digital Health 2020; 2: e303–13
  • “In conclusion, this study provided a proof of concept that CNN can accurately distinguish pancreatic cancer on portal venous CT images. The CNN model holds promise as a compute r­aided diagnostic tool to assist radiologists and clinicians in diagnosing pancreatic cancer.”
    Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation
    Kao-Lang Liu et al.
    Lancet Digital Health 2020; 2: e303–13
  • “Interpretation CNN could accurately distinguish pancreatic cancer on CT, with acceptable generalisability to images of patients from various races and ethnicities. CNN could supplement radiologist interpretation.”
    Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation
    Kao-Lang Liu et al.
    Lancet Digital Health 2020; 2: e303–13
  • “Deep learning is a type of machine learning method in which algorithms are trained to perform tasks by learning patterns from data rather than by explicit programming. Deep neural networks are inspired by biological neural networks and use a matrix of interconnected nodes to mimic the function of a biologic neuron. The basic unit of an artificial neural network is a node. It takes a set of input features, multiplies these features by corresponding weights in the form of mathematical equations, and then passes the output to the next layer of nodes. The deep network architecture uses multiple layers of interconnected nodes to develop a mathematical model that best fits the data. The outputs are compared with the “ground truth,” and errors are used as feedback to adjust the weights in the network to minimize error in subsequent iterations.”
    Pancreatic Cancer Imaging: A New Look at an Old Problem
    Linda C. Chu MD, Seyoun Park, Satomi Kawamoto, Alan L. Yuille , Ralph H. Hruban, Elliot K. Fishman
    Current Problems in Diagnostic Radiology (in press)

  • Automatic detection of pancreatic ductal adenocarcinoma (PDAC) with deep learning. (Left panel) Axial IV contrast- enhanced CT image shows a hypoenhancing mass in the pancreatic body (arrow) with dilated pancreatic duct (arrowhead). (Middle panel) Manual segmentation of the tumor (red), pancreatic duct (green), and background pancreas (blue). (Right panel) Deep network prediction of tumor (red), pancreatic duct (green) and background pancreas (blue).

  • Automatic detection of pancreatic neuroendocrine tumor (PanNET) with deep learning. (Left panel) Axial IV contrast-enhanced CT image shows a subtle hyperenhancing mass within the head of the pancreas (arrow). (Middle panel) Manual segmentation of tumor (pink) and background pancreas (blue). (Right panel) Deep network prediction of tumor (pink) and background pancreas (blue).

  • Automatic detection of intraductal papillary mucinous neoplasm (IPMN) with deep learning. (Left panel) Axial IV contrast-enhanced CT image shows multiple well-circumscribed cystic lesions in the pancreas (arrow). (Middle panel) Manual segmentation of cystic tumors (yellow) and background pancreas (blue). (Right panel) Deep network prediction of cystic tumors (yellow) and background pancreas (blue).

  • A schematic illustrating the radiomics feature extraction and analysis process. Radiomics features can be classified into signal intensity, shape, texture, and filtered features (e.g., wavelets and Laplacian of Gaussian [LoG]). (Left panel) Input of imaging datasets (normal vs. abnormal) with annotation of regions of interest. (Middle panel) Extraction of radiomics features, including histogram of voxel signal intensities, shape features based on surface rendering of region of interest, and filtered features. (Right panel) The raw data are processed through feature selection to identify the most relevant features. These features can be correlated with clinical outcomes in classification tasks.
  • “Radiomics features have also been used to predict PanNET grade, one of the most important prognostic factors in predicting patient survival. Qualitative features such as ill-defined margins, heterogeneous enhancement, low- level enhancement, vascular involvement, and main pancreatic duct dilatation have been reported to be helpful features in predicting higher tumor grade. Radiomics features achieved equivalent or superior performance compared to traditional clinical and imaging features. in most, but not all studies, with higher tumor grades in the majority of these studies, and with worse progression free survival. The addition of radiomics features to traditional CT features may improve the accuracy of PanNET grade prediction.”
    Pancreatic Cancer Imaging: A New Look at an Old Problem
    Linda C. Chu MD, Seyoun Park, Satomi Kawamoto, Alan L. Yuille , Ralph H. Hruban, Elliot K. Fishman
    Current Problems in Diagnostic Radiology (in press)
  • “Radiomics features have also been reported to be predictive of overall survival in patients with unresectable or locally advanced PDAC. Not surprisingly, the presence of metastatic disease at presentation was the most predictive of poor overall survival. factors. Radiomics features associated with tumor heterogeneity were also found to be poor prognostic factors. There is speculation that tumor hypoattenuation may reflect areas of hypoxic necrosis, which may suggest more aggressive underlying tumor biology as well as impaired response to chemotherapy and radiation therapy. Low attenuation may also be evidence of extensive venous invasion by the cancer.”
    Pancreatic Cancer Imaging: A New Look at an Old Problem
    Linda C. Chu MD, Seyoun Park, Satomi Kawamoto, Alan L. Yuille , Ralph H. Hruban, Elliot K. Fishman
    Current Problems in Diagnostic Radiology (in press)
  • "While VR uses a simple ray cast method to generate 3D images, CR uses Monte Carlo path tracing that takes direct and indirect illumination into account. With CR, each pixel is formed by thousands of rays passing through the volumetric dataset and includes effects of light rays from scatter and from voxels adjacent to the paths of the rays. CR has the potential to more accurately depict complex anatomy. When applied to pancreatic imaging, CR can be used to accentuate focal textural change and enhance appreciation of internal architecture (e.g., septations, mural nodules) to improve their visualization and assist in tumor classification.”
    Pancreatic Cancer Imaging: A New Look at an Old Problem
    Linda C. Chu MD, Seyoun Park, Satomi Kawamoto, Alan L. Yuille , Ralph H. Hruban, Elliot K. Fishman
    Current Problems in Diagnostic Radiology (in press)
  • “Augmented reality (AR) is another advanced visualization technique that may improve treatment planning as well as intraoperative navigation. AR can superimpose holographic representations of imaging data onto the real-world environment through the use of handheld displays or head-mounted see-through glasses. Preliminary studies on AR applications in pancreatic surgery have shown that these holographic images may be helpful in proper selection of resection margin and in defining the spatial relationship between the tumor and adjacent organs and vasculature. AR surgical navigation may be particularly valuable during laparoscopic or robotic-assisted surgery due to limited visualization and tactile feedback during surgery.”
    Pancreatic Cancer Imaging: A New Look at an Old Problem
    Linda C. Chu MD, Seyoun Park, Satomi Kawamoto, Alan L. Yuille , Ralph H. Hruban, Elliot K. Fishman
    Current Problems in Diagnostic Radiology (in press)
  • “Although radiomics has the potential to provide personalized imaging biomarkers for risk stratification and prognostication, there are currently no standards for image acquisition or feature extraction. Published studies differ regarding image acquisition, segmentation, and the types and numbers of radiomics features that were extracted. Each of these factors can affect the reproducibility of radiomics signatures. Although some of this variability may be mitigated through image compensation methods,86 further work is needed to define the optimal image acquisition and feature extraction protocols. While these preliminary studies appear promising, many of them lack internal and external validation to ensure the generalizability of the results. Several studies also lack head-to-head comparisons between radiomics and expert radiologists to demonstrate the incremental clinical benefit of radiomics as opposed to current standard of care. The potential of advanced visualization techniques in guiding patient management has been explored in small single-center case-series, and these results also require further validation.”
    Pancreatic Cancer Imaging: A New Look at an Old Problem
    Linda C. Chu MD, Seyoun Park, Satomi Kawamoto, Alan L. Yuille , Ralph H. Hruban, Elliot K. Fishman
    Current Problems in Diagnostic Radiology (in press)
  • “Pancreatic ductal adenocarcinoma (PDAC) segmentation is one of the most challenging tumor segmentation tasks, yet critically important for clinical needs. Previous work on PDAC segmentation is limited to the moderate amounts of annotated patient images (n<300) from venous or venous+arterial phase CT scans. Based on a new self-learning framework, we propose to train the PDAC segmentation model using a much larger quantity of patients (n≈1,000), with a mix of annotated and un- annotated venous or multi-phase CT images. Pseudo annotations are generated by combining two teacher models with different PDAC segmentation specialties on unannotated images, and can be further refined by a teaching assistant model that identifies associated vessels around the pancreas.”
    Robust Pancreatic Ductal Adenocarcinoma Segmentation with Multi-Institutional Multi-Phase Partially-Annotated CT Scans
    Ling Zhang et al.
    arXiv: August 2020 (in press)
  • “Fully automated and accurate segmentation of pancreatic ductal adenocarcinoma (PDAC) is one of the most challenging tumor segmentation tasks, in the aspects of complex abdominal structures, large variations in morphology and appearance, low image contrast and fuzzy/uncertain boundary, etc. Previous studies introduce the cascade UNet for segmenting venous phase CT and hyperpairing network for segmenting venous+arterial phases CT and achieving mean Dice scores of 0.52 and 0.64, respectively. By incorporating nnUNet into a new self-learning framework with two teachers and one teaching assistant to segment three-phases of CT scans, our method reaches a Dice coefficient of 0.71, similar to the inter-observer variability between radiologists. This provides promise that a radiologist-level performance for accurate PDAC tumor segmentation in multi-phase CT imaging can be achieved through our computerized method.”
    Robust Pancreatic Ductal Adenocarcinoma Segmentation with Multi-Institutional Multi-Phase Partially-Annotated CT Scans
    Ling Zhang et al.
    arXiv: August 2020 (in press)

  • Robust Pancreatic Ductal Adenocarcinoma Segmentation with Multi-Institutional Multi-Phase Partially-Annotated CT Scans
    Ling Zhang et al.
    arXiv: August 2020 (in press)
  • Background: To identify preoperative computed tomography radiomics texture features which correlate with resection margin status and prognosis in resected pancreatic head adenocarcinoma.
    Methods: Improved prognostication methods utilizing novel non-invasive radiomic techniques may accurately predict resection margin status preoperatively. In an ongoing concerning pancreatic head adenocarcinoma, the venous enhanced CT images of 86 patients who underwent pancreaticoduodenectomy were selected, and the resection margin (>1 mm or ≤1 mm) was identified by pathological examination. Three regions of interests (ROIs) were then taken from superior to inferior facing the superior mesenteric vein and artery. Subsequent Laplacian-Dirichlet based texture analysis methods extracting algorithm of texture features within ROIs were analyzed and assessed in relation to patient prognosis.
    Results: Patients with >1 mm resection margin had an overall improved survival compared to ≤1 mm (P < 0.05). Distance 1 and 2 of Gray level co-occurrence matrix, high Gray-level run emphasis of run-length matrix and average of wavelet transform (all P < 0.05) were correlated with resection margin status (Area under the curve was 0.784, sensitivity was 75% and specicity was 79%). The energy of wavelet transform, the measure of smoothness of histogram and the variance in 2 direction of Gabor transform are independent predictors of overall survival prognosis, independent of resection margin.
    Conclusions: Resection margin status (>1 mm vs ≤1 mm) is a key prognostic factor in pancreatic adenocarcinoma and CT radiomic analysis have the potential to predict resection margin status preoperatively, and the radiomic labels may improve selection neoadjucant therapy.
    Radiomics Signatures of Computed Tomography Imaging for Predicting Resection Margin Status in Pancreatic Head Adenocarcinoma
    Jinheng Liu et al.
    BMC Surgery (in press)
  • Results: Patients with >1 mm resection margin had an overall improved survival compared to ≤1 mm (P < 0.05). Distance 1 and 2 of Gray level co-occurrence matrix, high Gray-level run emphasis of run-length matrix and average lter of wavelet transform (all P < 0.05) were correlated with resection margin status (Area under the curve was 0.784, sensitivity was 75% and specificity was 79%). The energy of wavelet transform, the measure of smoothness of histogram and the variance in 2 direction of Gabor transform are independent predictors of overall survival prognosis, independent of resection margin.
    Conclusions: Resection margin status (>1 mm vs ≤1 mm) is a key prognostic factor in pancreatic adenocarcinoma and CT radiomic analysis have the potential to predict resection margin status preoperatively, and the radiomic labels may improve selection neoadjucant therapy.
    Radiomics Signatures of Computed Tomography Imaging for Predicting Resection Margin Status in Pancreatic Head Adenocarcinoma
    Jinheng Liu et al.
    BMC Surgery (in press)
  • Background: To identify preoperative computed tomography radiomics texture features which correlate with resection margin status and prognosis in resected pancreatic head adenocarcinoma.
    Methods: Improved prognostication methods utilizing novel non-invasive radiomic techniques may accurately predict resection margin status preoperatively. In an ongoing concerning pancreatic head adenocarcinoma, the venous enhanced CT images of 86 patients who underwent pancreaticoduodenectomy were selected, and the resection margin (>1 mm or ≤1 mm) was identified by pathological examination. Three regions of interests (ROIs) were then taken from superior to inferior facing the superior mesenteric vein and artery. Subsequent Laplacian-Dirichlet based texture analysis methods extracting algorithm of texture features within ROIs were analyzed and assessed in relation to patient prognosis.
    Radiomics Signatures of Computed Tomography Imaging for Predicting Resection Margin Status in Pancreatic Head Adenocarcinoma
    Jinheng Liu et al.
    BMC Surgery (in press)
  • “Radiomic texture analysis of pre-operative enhanced CT images can be used for accurate preoperative assessment of resection margins in patients with pancreatic ahead adenocarcinoma providing clinicians alongside patients a more non-invasive means of perioperative prognostication to guide management.”
    Radiomics Signatures of Computed Tomography Imaging for Predicting Resection Margin Status in Pancreatic Head Adenocarcinoma
    Jinheng Liu et al.
    BMC Surgery (in press)

  • Radiomics Signatures of Computed Tomography Imaging for Predicting Resection Margin Status in Pancreatic Head Adenocarcinoma
    Jinheng Liu et al.
    BMC Surgery (in press)

  • Radiomics Signatures of Computed Tomography Imaging for Predicting Resection Margin Status in Pancreatic Head Adenocarcinoma
    Jinheng Liu et al.
    BMC Surgery (in press)
  • “PDAC is the most common pancreatic malig- nancy, accounting for more than 85% of pancreatic tumors. It is typically a disease of elderly patients, with a mean age at presentation of 68 years and a male-to-female ratio of 1.6:1. After colorectal cancer, it is the second most common cancer of the digestive system in the United States, and its incidence is rising sharply.The development of pancreatic cancer is strongly related to smoking, family history, obesity, long-standing diabetes, and chronic pancreatitis. Early stages of PDAC are clinically silent. Abdominal pain is the most frequently reported clinical symptom, even when the tumor is small (<2 cm).”
    Pancreatic Ductal Adenocarcinoma and Its Variants: Pearls and Perils
    Schawkat K et al.
    RadioGraphics 2020; 40:0000–0000
  • "With the development of AI and all its potential wonders in terms of increasing the accuracy of our diagnostic capabilities and potentially improving patient care, we must also be concerned about the potential dark side by bad actors. The sooner organized radiology and organized medicine address these issues with clarity, the more stable and protected the health care system and our patients will be from those intent on creating harm and havoc by abusing AI. The acceleration of data sharing during the current pandemic exposes critical vulnerabilities in data security. It reminds us of the pervasive threat that bad actors can and will exploit any technology for their selfish gains. Doing nothing is not a viable strategy, but acting in a concerted effort will lead us to the protection we need and is important as we push AI development over the next several years.”
    The Potential Dangers of Artificial Intelligence for Radiology and Radiologists
    Linda C. Chu, MD, Anima Anandkumar, PhD, Hoo Chang Shin, PhD, Elliot K. Fishman, MD
    JACR (in press)
  • “Pancreatic cancer continues to be one of the deadliest malignancies and is the third leading cause of cancer-related mortality in the United States. Based on several models, it is projected to become the second leading cause of cancer-related deaths by 2030. Although the overall survival rate for patients diagnosed with pancreatic cancer is less than 10%, survival rates are increasing in those whose cancers are detected at an early stage, when intervention is possible. There are, however, no reli- able biomarkers or imaging technology that can detect early-stage pancreatic cancer or accurately identify precursors that are likely to progress to malignancy.”
    Prediagnostic Image Data, Artificial Intelligence, and Pancreatic Cancer A Tell-Tale Sign to Early Detection
    Matthew R. Young et al.
    Pancreas 2020;49: 882–886
  • "The challenge now is to develop imaging biomarkers and models that can further improve sensitivity for the detection of early-stage PDACs and aggressive neoplasms while mitigating diagnostic uncertainty in evaluation of premalignant abnormalities. Augmented reality, artificial intelligence (AI), and related computa- tional techniques can uncover these subtle patterns, improve image interpretation, and streamline diagnostic workflows.”
    Prediagnostic Image Data, Artificial Intelligence, and Pancreatic Cancer A Tell-Tale Sign to Early Detection
    Matthew R. Young et al.
    Pancreas 2020;49: 882–886
  • "Currently, identification of localized pancreatic cancer is mostly incidental as localized pancreatic cancer is asymptomatic. What is urgently needed are minimally invasive screening strategies with a high clinical sensitivity and specificity to identity early-stage cancer and improve these grim statistics. To this end, it is particularly important to develop tests that have high specificity because a false-positive test may trigger unnecessary invasive procedures, which add their own risk of morbidity and mortality.”
    Prediagnostic Image Data, Artificial Intelligence, and Pancreatic Cancer A Tell-Tale Sign to Early Detection
    Matthew R. Young et al.
    Pancreas 2020;49: 882–886
  • There are many challenges that need to be mitigated in the development of an image repository to enable AI system development. These include the following:
    (1) What are the requirements for defining image annotation? 
    (2) What are the main concerns with depositing patient imaging data?
    (3) What are the definitions of an AI-specific clinical use cases?
    (4) What are the benefits and drawbacks of alternative data sharing in facilitating AI development? 

  • Prediagnostic Image Data, Artificial Intelligence, and Pancreatic Cancer A Tell-Tale Sign to Early Detection
    Matthew R. Young et al.
    Pancreas 2020;49: 882–886

  • Prediagnostic Image Data, Artificial Intelligence, and Pancreatic Cancer A Tell-Tale Sign to Early Detection
    Matthew R. Young et al.
    Pancreas 2020;49: 882–886
  • “The AI-driven diagnostic software has the potential to trans- form early detection of pancreatic cancer by improving accuracy and consistency of interpretation of radiologic imaging scans and related patient data. The development of reproducible AI systems requires access to current, large, diverse, and multisite data sets, which are subject to numerous data sharing limitations. Future efforts are likely to involve alternative data sharing solutions to enable the development of both public and private AI-ready data resources. Early detection of pancreatic cancer represents an attractive AI use case, well matched to benefit from the MTD challenge approach. This approach will significantly expand the use of sensitive data to improve early detection of pancreatic cancer and lay the foundation for the development of federated architectures for real-world medical data in general.”
    Prediagnostic Image Data, Artificial Intelligence, and Pancreatic Cancer A Tell-Tale Sign to Early Detection
    Matthew R. Young et al.
    Pancreas 2020;49: 882–886
  • Purpose: The purpose of this study was to determine whether computed tomography (CT)-based machine learning of radiomics features could help distinguish autoimmune pancreatitis (AIP) from pancreatic ductal adenocarcinoma (PDAC).
    Results: The pancreas was diffusely involved in 37 (37/89; 41.6%) patients with AIP and not diffusely in 52(52/89; 58.4%) patients. Using machine learning, 95.2% (59/62; 95% confidence interval [CI]: 89.8–100%),83.9% (52:67; 95% CI: 74.7–93.0%) and 77.4% (48/62; 95% CI: 67.0–87.8%) of the 62 test patients werecorrectly classified as either having PDAC or AIP with thin-slice venous phase, thin-slice arterial phase, and thick-slice venous phase CT, respectively. Three of the 29 patients with AIP (3/29; 10.3%) were incorrectly classified as having PDAC but all 33 patients with PDAC (33/33; 100%) were correctly classified with thin-slice venous phase with 89.7% sensitivity (26/29; 95% CI: 78.6–100%) and 100% specificity (33/33;95% CI: 93–100%) for the diagnosis of AIP, 95.2% accuracy (59/62; 95% CI: 89.8–100%) and area under the curve of 0.975 (95% CI: 0.936–1.0).
    Conclusions: Radiomic features help differentiate AIP from PDAC with an overall accuracy of 95.2%.
  • Purpose: The purpose of this study was to determine whether computed tomography (CT)-based machine learning of radiomics features could help distinguish autoimmune pancreatitis (AIP) from pancreatic ductal adenocarcinoma (PDAC).
    Conclusions: Radiomic features help differentiate AIP from PDAC with an overall accuracy of 95.2%.
    Differentiating autoimmune pancreatitis from pancreatic ductal adenocarcinoma with CT radiomics features
    S. Park, L.C. Chu, R.H. Hruban, Vogelstein, K.W. Kinzler, A.L. Yuille, Fouladi, S. Shayesteh, S. Ghandili, C.L. Wolfgang, R. Burkhart, J. He, E.K. Fishman, S. Kawamoto
    Diagnostic and Interventional Imaging (in press)
  • •CT radiomics differentiates AIP from PDAC with 89.7% sensitivity and 100% specificity.
    •Thin slice CT radiomics better differentiates AIP from PDAC than thick slice CT radiomics.
    •Venous phase CT radiomics better differentiates AIP from PDAC than arterial phase radiomics.
    Differentiating autoimmune pancreatitis from pancreatic ductal adenocarcinoma with CT radiomics features
    S. Park, L.C. Chu, R.H. Hruban, Vogelstein, K.W. Kinzler, A.L. Yuille, Fouladi, S. Shayesteh, S. Ghandili, C.L. Wolfgang, R. Burkhart, J. He, E.K. Fishman, S. Kawamoto
    Diagnostic and Interventional Imaging (in press)
  • “AIP has clinical and imaging features that overlap with those of pancreatic ductal adenocarcinoma (PDAC) and can pose a significant diagnostic dilemma even for experienced radiologists . The management of these two conditions is markedly different. Patients with AIP are initially treated with oral corticosteroids, while patients with PDAC are treated with a combination of surgical resection and chemotherapy. The most common presentation of AIP is obstructive jaundice and pancreatic enlargement, which mimics that of PDAC and 2–6% of patients undergoing surgical resection for suspected pancreatic cancer are actually diagnosed with AIP upon histopathological analysis. Computed tomography (CT) plays an important role in the evaluation of suspected pancreatic cancer, and is often the initial diagnostic imaging modality. It is of utmost importance to correctly differentiate AIP from PDAC early in the disease process so as to administer the proper treatment and avoid unnecessary pancreatic resections in patients with AIP.”
    Differentiating autoimmune pancreatitis from pancreatic ductal adenocarcinoma with CT radiomics features
    S. Park, L.C. Chu, R.H. Hruban, Vogelstein, K.W. Kinzler, A.L. Yuille, Fouladi, S. Shayesteh, S. Ghandili, C.L. Wolfgang, R. Burkhart, J. He, E.K. Fishman, S. Kawamoto
    Diagnostic and Interventional Imaging (in press)

  • Differentiating autoimmune pancreatitis from pancreatic ductal adenocarcinoma with CT radiomics features
    S. Park, L.C. Chu, R.H. Hruban, Vogelstein, K.W. Kinzler, A.L. Yuille, Fouladi, S. Shayesteh, S. Ghandili, C.L. Wolfgang, R. Burkhart, J. He, E.K. Fishman, S. Kawamoto
    Diagnostic and Interventional Imaging (in press)
  • "In conclusion, radiomics analysis of CT images is reasonably accurate in differentiating AIP from PDAC. Using such features, in combination with clinical and standard radiologic analyses, may improve the accuracy of AID diagnosis and spare patients’ unnecessary surgical procedure.”
    Differentiating autoimmune pancreatitis from pancreatic ductal adenocarcinoma with CT radiomics features
    S. Park, L.C. Chu, R.H. Hruban, Vogelstein, K.W. Kinzler, A.L. Yuille, Fouladi, S. Shayesteh, S. Ghandili, C.L. Wolfgang, R. Burkhart, J. He, E.K. Fishman, S. Kawamoto
    Diagnostic and Interventional Imaging (in press)
  • "Our results showed that by combining radiomics features, AIP could be distinguished from PDAC with a sensitivity of 89.7% and a specificity of 100%, and an overall accuracy of 95.2%. Among 3 patients with focal AIP were falsely classified as PDAC using radiomics features, two patients had focal AIP in the head with a plastic stent in the common bile duct, which can sensitively affect to the quantitative feature computation. In our study, the accuracy was higher than that in a previous study that evaluated CT to differentiate AIP from PDAC based on morphological features. In that study, the mean accuracies for diagnosing AIP and PDAC were 68% and 83%, respectively. In our study, AIP was considered as a diagnosis or differential diagnosis by the radiologists in only in 67% of patients with AIP not already suspected to be AIP at the time of CT examination.”
    Differentiating autoimmune pancreatitis from pancreatic ductal adenocarcinoma with CT radiomics features
    S. Park, L.C. Chu, R.H. Hruban, Vogelstein, K.W. Kinzler, A.L. Yuille, Fouladi, S. Shayesteh, S. Ghandili, C.L. Wolfgang, R. Burkhart, J. He, E.K. Fishman, S. Kawamoto
    Diagnostic and Interventional Imaging (in press)
  • “We found that radiomics features were better at distinguishing AIP from PDAC using venous phase CT images than using arterial phase images. We also performed radiomics analysis on both thin- and thick-slice reconstructions. We found that thin-slice CT based radiomics signature had better diagnostic performance than thick-slice, as reported in pulmonary nodules and lung cancer in prior studies.”
    Differentiating autoimmune pancreatitis from pancreatic ductal adenocarcinoma with CT radiomics features
    S. Park, L.C. Chu, R.H. Hruban, Vogelstein, K.W. Kinzler, A.L. Yuille, Fouladi, S. Shayesteh, S. Ghandili, C.L. Wolfgang, R. Burkhart, J. He, E.K. Fishman, S. Kawamoto
    Diagnostic and Interventional Imaging (in press)
  • Purpose: The purpose of this study is to evaluate diagnostic performance of a commercially available radiomics research prototype vs. an in-house radiomics software in the binary classification of CT images from patients with pancreatic ductal adenocarcinoma (PDAC) vs. healthy controls.
    Conclusion: Commercially available and in-house radiomics software achieve similar diagnostic performance, which may lower the barrier of entry for radiomics research and allow more clinician-scientists to perform radiomics research.
    Diagnostic performance of commercially available vs. in‐house radiomics software in classification of CT images from patients with pancreatic ductal adenocarcinoma vs. healthy controls
    Linda C. Chu · Berkan Solmaz · Seyoun Park · Satomi Kawamoto · Alan L. Yuille · Ralph H. Hruban · Elliot K. Fishman
    Abdominal Radiology https://doi.org/10.1007/s00261-020-02556-w
  • “Results: When 40 radiomics features were used in the random forest classification, in-house software achieved superior sensitivity (1.00) and accuracy (0.992) compared to the commercially available research prototype (sensitivity = 0.950, accuracy = 0.968). When the number of features was reduced to five features, diagnostic performance of the in-house soft- ware decreased to sensitivity (0.950), specificity (0.923), and accuracy (0.936). Diagnostic performance of the commercially available research prototype was unchanged.”
    Diagnostic performance of commercially available vs. in‐house radiomics software in classification of CT images from patients with pancreatic ductal adenocarcinoma vs. healthy controls
    Linda C. Chu · Berkan Solmaz · Seyoun Park · Satomi Kawamoto · Alan L. Yuille · Ralph H. Hruban · Elliot K. Fishman
    Abdominal Radiology https://doi.org/10.1007/s00261-020-02556-w
  • “Radiomics has the potential to generate imaging biomarkers for classification and prognostication. Technical parameters from image acquisition to feature extraction and analysis have the potential to affect radiomics features. The current study used the same CT images with manual segmentation on both a commercially available research prototype and in-house radiomics software to control for any variability at the image acquisition step and compared the diagnostic performance of the two programs. Both programs achieved similar diagnostic performance in the binary classification of CT images from patients with PDAC and healthy control subjects, despite differences in the radiomics fea-tures they employed (854 features in commercial program vs. 478 features in in-house program).”
    Diagnostic performance of commercially available vs. in‐house radiomics software in classification of CT images from patients with pancreatic ductal adenocarcinoma vs. healthy controls
    Linda C. Chu · Berkan Solmaz · Seyoun Park · Satomi Kawamoto · Alan L. Yuille · Ralph H. Hruban · Elliot K. Fishman
    Abdominal Radiology https://doi.org/101007/s00261-020-02556-w
  • "This is reassuring that even though there may be variations in the computed values for radiomics features, the differences do not seem to significantly impact the overall diagnostic performance of the constellation of radiomics features. This is important for the broader implementation of radiomics research. Currently, many radiomics studies have been performed using proprietary in-house software, which requires in-house expertise in computer science, a luxury that only a few academic centers can afford. The results of this study show that commercially available radiomics software may be a viable alternative to in-house computer science expertise, which can lower the barrier of entry for radiomics research and allow clinicians to validate findings of the published studies with their own local datasets.”
    Diagnostic performance of commercially available vs. in‐house radiomics software in classification of CT images from patients with pancreatic ductal adenocarcinoma vs. healthy controls
    Linda C. Chu · Berkan Solmaz · Seyoun Park · Satomi Kawamoto · Alan L. Yuille · Ralph H. Hruban · Elliot K. Fishman
    Abdominal Radiology https://doi.org/101007/s00261-020-02556-w
  • “This study showed that a commercially available radiomics software may be able to achieve similar diagnostic performance as an in-house radiomics software. The results obtained from one radiomics software may be transferrable to another system. Availability of commercial radiom ics software may lower the barrier of entry for radiomics research and allow more researchers to engage in this exciting area of research.”
    Diagnostic performance of commercially available vs. in‐house radiomics software in classification of CT images from patients with pancreatic ductal adenocarcinoma vs. healthy controls
    Linda C. Chu · Berkan Solmaz · Seyoun Park · Satomi Kawamoto · Alan L. Yuille · Ralph H. Hruban · Elliot K. Fishman
    Abdominal Radiology https://doi.org/101007/s00261-020-02556-w
  • “Accurate and robust segmentation of abdominal organs on CT is essential for many clinical applications such as computer-aided diagnosis and computer-aided surgery. But this task is challenging due to the weak boundaries of organs, the complexity of the background, and the variable sizes of different organs. To address these challenges, we introduce a novel framework for multi-organ segmentation of abdominal regions by using organ-attention networks with reverse connections (OAN-RCs) which are applied to 2D views, of the 3D CT volume, and output estimates which are combined by statistical fusion exploiting structural similarity. More specifically, OAN is a two-stage deep convolutional network, where deep net- work features from the first stage are combined with the original image, in a second stage, to reduce the complex background and enhance the discriminative information for the target organs.”
    Abdominal multi-organ segmentation with organ-attention networks and statistical fusion
    Wang Y, Zhou Y, Shen W, Park S, Fishman EK, Yuille AL.
    Med Image Anal. 2019 Jul;55:88-102. doi: 10.1016/j.media.2019.04.005. Epub 2019 Apr 18.
  • "First, many abdominal organs have weak boundaries between spatially adjacent structures on CT, e.g. between the head of the pancreas and the duodenum. In addition, the entire CT volume includes a large variety of different complex structures. Morpho- logical and topological complexity includes anatomically connected structures such as the gastrointestinal (GI) track (stomach, duodenum, small bowel and colon) and vascular structures. The correct anatomical borders between connected structures may not be always visible in CT, especially in sectional images (i.e., 2D slices), and may be indicated only by subtle texture and shape change, which causes uncertainty even for human experts. This makes it hard for deep networks to distinguish the target organs from the complex background.”
    Abdominal multi-organ segmentation with organ-attention networks and statistical fusion
    Wang Y, Zhou Y, Shen W, Park S, Fishman EK, Yuille AL.
    Med Image Anal. 2019 Jul;55:88-102. doi: 10.1016/j.media.2019.04.005. Epub 2019 Apr 18.
  • “In general, 3D deep networks face far greater complex challenges than 2D deep networks. Both approaches rely heavily on graphics processing units (GPUs) but these GPUs have limited memory size which makes it difficult when dealing with full 3D CT volumes compared to 2D CT slices (which require much less memory). In addition, 3D deep networks typically require many more parameters than 2D deep networks and hence require much more training data, unless they are re- stricted to patches.”
    Abdominal multi-organ segmentation with organ-attention networks and statistical fusion
    Wang Y, Zhou Y, Shen W, Park S, Fishman EK, Yuille AL.
    Med Image Anal. 2019 Jul;55:88-102. doi: 10.1016/j.media.2019.04.005. Epub 2019 Apr 18.

  • Abdominal multi-organ segmentation with organ-attention networks and statistical fusion
    Wang Y, Zhou Y, Shen W, Park S, Fishman EK, Yuille AL.
    Med Image Anal. 2019 Jul;55:88-102. doi: 10.1016/j.media.2019.04.005. Epub 2019 Apr 18.
  • “In this paper, we proposed a novel framework for multi- organ segmentation using OAN-RCs with statistical fusion exploit- ing structural similarity. Our two-stage organ-attention network reduces uncertainties at weak boundaries, focuses attention on or- gan regions with simple context, and adjusts FCN error by training the combination of original images and OAMs. Reverse connections deliver abstract level semantic information to lower layers so that hidden layers can be assisted to contain more semantic information and give good results even for small organs.”
    Abdominal multi-organ segmentation with organ-attention networks and statistical fusion
    Wang Y, Zhou Y, Shen W, Park S, Fishman EK, Yuille AL.
    Med Image Anal. 2019 Jul;55:88-102. doi: 10.1016/j.media.2019.04.005. Epub 2019 Apr 18.

  • Abdominal multi-organ segmentation with organ-attention networks and statistical fusion
    Wang Y, Zhou Y, Shen W, Park S, Fishman EK, Yuille AL.
    Med Image Anal. 2019 Jul;55:88-102. doi: 10.1016/j.media.2019.04.005. Epub 2019 Apr 18.

  • Abdominal multi-organ segmentation with organ-attention networks and statistical fusion
    Wang Y, Zhou Y, Shen W, Park S, Fishman EK, Yuille AL.
    Med Image Anal. 2019 Jul;55:88-102. doi: 10.1016/j.media.2019.04.005. Epub 2019 Apr 18.
  • “In addition to traditional methods, cinematic rendering (CR) as a novel 3D rendering technique can be used to generate photorealistic with more accurate information regarding the anatomical details. CR can assist clinicians to visualize precisely the extent of tumor vascular invasion, which might be critical for surgical planning; however, the feasibility of this method and other novel techniques in routine clinical practice is yet to be studied.”
    Pitfalls in the MDCT of pancreatic cancer: strategies for minimizing errors
    Arya Haj‐Mirzaian · Satomi Kawamoto · Atif Zaheer · Ralph H. Hruban · Elliot K. Fishman · Linda C. Chu
    Abdominal Radiology 2020 (in press)
  • Purpose: The purpose of this study was to report procedures developed to annotate abdominal computed tomography (CT) images from subjects without pancreatic disease that will be used as the input for deep convolutional neural networks (DNN) for development of deep learning algorithms for automatic recognition of a normal pancreas.
    Results: A total of 1150 dual-phase CT datasets from 575 subjects were annotated. There were 229 men and 346 women (mean age: 45 ± 12 years; range: 18—79 years). The mean intra- observer intra-subject dual-phase CT volume difference of all annotated structures was 4.27 mL (7.65%). The deep network prediction for multi-organ segmentation showed high fidelity with 89.4% and 1.29 mm in terms of mean Dice similarity coefficients and mean surface distances, respectively.
    Annotated normal CT data of the abdomen for deep learning: Challenges and strategies for implementation
    S. Park, L.C. Chu, E.K. Fishman, A.L. Yuille, B. Vogelstein,, K.W. Kinzler et al
    Diagn Interv Imaging. 2020 Jan;101(1):35-44.
  • “Conclusions: A reliable data collection/annotation process for abdominal structures was devel- oped. This process can be used to generate large datasets appropriate for deep learning.”
    Annotated normal CT data of the abdomen for deep learning: Challenges and strategies for implementation
    S. Park, L.C. Chu, E.K. Fishman, A.L. Yuille, B. Vogelstein,, K.W. Kinzler et al
    Diagn Interv Imaging. 2020 Jan;101(1):35-44.
  • Annotated normal CT data of the abdomen for deep learning: Challenges and strategies for implementation S. Park, L.C. Chu, E.K. Fishman, A.L. Yuille, B. Vogelstein,, K.W. Kinzler et al Diagn Interv Imaging. 2020 Jan;101(1):35-44.
  • Annotated normal CT data of the abdomen for deep learning: Challenges and strategies for implementation
    S. Park, L.C. Chu, E.K. Fishman, A.L. Yuille, B. Vogelstein,, K.W. Kinzler et al
    Diagn Interv Imaging. 2020 Jan;101(1):35-44.

  • “In conclusion, we developed a reliable and unique data collection and annotation process for abdominal structures using volumetric CT. The collected data can be used to train the deep learning network for automated recognition of normal abdominal organs. The success of this effort was dependent on a multidisciplinary team including radiologists, computer scientists, oncologists, and pathologists that have worked closely together. Pathologists confirmed that the pancreas in all subjects were normal without pancreatic neoplasms or other pathology. Oncologists provided expert guidance in experimental deign and data analysis.”
    Annotated normal CT data of the abdomen for deep learning: Challenges and strategies for implementation
    S. Park, L.C. Chu, E.K. Fishman, A.L. Yuille, B. Vogelstein,, K.W. Kinzler et al
    Diagn Interv Imaging. 2020 Jan;101(1):35-44.

  • Assessing Radiology Research on Artificial Intelligence:
    A Brief Guide for Authors, Reviewers, and Readers—From the Radiology Editorial Board
    David A.Bluemke et al.
    Radiology 2019; (in press) https://doi.org/10.1148/radiol.2019192515

  • Application of Deep Learning to Pancreatic Cancer Detection: Lessons Learned From Our Initial Experience.
    Chu LC, Park S, Kawamoto S, Wang Y, Zhou Y, Shen W, Zhu Z, Xia Y, Xie L, Liu F, Yu Q, Fouladi DF, Shayesteh S, Zinreich E, Graves JS, Horton KM, Yuille AL, Hruban RH, Kinzler KW, Vogelstein B, Fishman EK.
    J Am Coll Radiol. 2019 Sep;16(9 Pt B):1338-1342
  • “There is a common perception that one can simply provide any number of unprocessed cases to the computer, and AI can then easily perform the discovery or classification task. This approach is referred to as unsupervised learning, in which the deep-learning algorithm is presented with unlabeled data and learns to group the data by similarities or differences. Although this approach is plausible, complex image analysis, such as the detection of pancreatic cancer, may require supervised learning to achieve acceptable results.”
    Application of Deep Learning to Pancreatic Cancer Detection: Lessons Learned From Our Initial Experience.
    Chu LC, Park S, Kawamoto S, et al
    J Am Coll Radiol. 2019 Sep;16(9 Pt B):1338-1342
  • In supervised learning, the algorithm is provided with labeled data, referred to as ground truth, which is used as feedback to improve the algorithm during each iteration. The degree of data labeling can range from a per case level of normal versus abnormal to more detailed labeling in which the boundaries of each region of interest are drawn on the image on every image slice; this boundary drawing is referred to as “segmentation.” Because we have chosen to tackle a difficult AI application, we decided that supervised learning with high- quality input data would yield the best chance of success.
    Application of Deep Learning to Pancreatic Cancer Detection: Lessons Learned From Our Initial Experience.
    Chu LC, Park S, Kawamoto S, et al
    J Am Coll Radiol. 2019 Sep;16(9 Pt B):1338-1342

  • Application of Deep Learning to Pancreatic Cancer Detection: Lessons Learned From Our Initial Experience.
    Chu LC, Park S, Kawamoto S, et al
    J Am Coll Radiol. 2019 Sep;16(9 Pt B):1338-1342

  • Application of Deep Learning to Pancreatic Cancer Detection: Lessons Learned From Our Initial Experience.
    Chu LC, Park S, Kawamoto S, et al
    J Am Coll Radiol. 2019 Sep;16(9 Pt B):1338-1342
  • “Our initial decision to train the deep network to recognize all major abdominal organs instead of focusing on the pancreas proved to be a wise investment of time and resources. As we reviewed the false positives, the deep network occasionally predicted the duodenum or jejunum as an exophytic tumor. This was especially problematic in thin patients with poor fat planes. As we trained the deep network to recognize and segment the major abdominal organs, we were able to use this algorithm to prune out false- positive predictions that overlapped with other organs.”
    Application of Deep Learning to Pancreatic Cancer Detection: Lessons Learned From Our Initial Experience.
    Chu LC, Park S, Kawamoto S, et al
    J Am Coll Radiol. 2019 Sep;16(9 Pt B):1338-1342
  • “In the future, we envision that that AI system for automatic PDAC detection will be seamlessly integrated into the radiology workflow as a “second reader,” similar to how computer-aided diagnosis operates in mammographic screening. The AI system will directly receive the CT data sets from the PACS, automatically segment the abdominal organs, and annotate any suspicious pancreatic pathology. These annotated cases will be sent back to the PACS for the radiologist to review. The “second reader” can improve diagnostic confidence and has the potential to identify subtle cases that can be missed by a busy radiologist. By increasing the sensitivity and accuracy of PDAC detection, AI- integrated workflow has the potential to significantly improve patient outcomes. As radiologists, we should not sit on the sidelines. Instead, we should actively engage the AI revolution, hoping to enhance our efficiency and reduce our errors, eventually improving patient outcomes.”
    Application of Deep Learning to Pancreatic Cancer Detection: Lessons Learned From Our Initial Experience.
    Chu LC, Park S, Kawamoto S, et al
    J Am Coll Radiol. 2019 Sep;16(9 Pt B):1338-1342
  • “In the future, we envision that that AI system for automatic PDAC detection will be seamlessly integrated into the radiology workflow as a “second reader,” similar to how computer-aided diagnosis operates in mammographic screening. The AI system will directly receive the CT data sets from the PACS, automatically segment the abdominal organs, and annotate any suspicious pancreatic pathology. These annotated cases will be sent back to the PACS for the radiologist to review. The “second reader” can improve diagnostic confidence and has the potential to identify subtle cases that can be missed by a busy radiologist.”
    Application of Deep Learning to Pancreatic Cancer Detection: Lessons Learned From Our Initial Experience.
    Chu LC, Park S, Kawamoto S, et al
    J Am Coll Radiol. 2019 Sep;16(9 Pt B):1338-1342
  • “We aim at segmenting a wide variety of organs, including tiny targets (e.g., adrenal gland) and neoplasms (e.g., pancreatic cyst), from abdominal CT scans. This is a challenging task in two aspects. First, some organs (e.g., the pancreas), are highly variable in both anatomy and geometry, and thus very difficult to depict. Second, the neoplasms often vary a lot in its size, shape, as well as its location within the organ. Third, the targets (organs and neoplasms) can be considerably small compared to the human body, and so standard deep networks for segmentation are often less sensitive to these targets and thus predict less accurately especially around their boundaries.”
    Recurrent Saliency Transformation Network for Tiny Target Segmentation in Abdominal CT Scans
    Lingxi Xie, Qihang Yu, Yan Wang, Yuyin Zhou, Elliot K. Fishman, and Alan L. Yuille
    IEEE Trans Med Imaging. 2019 Jul 23. doi: 10.1109/TMI.2019.2930679
  • In this paper, we present an end-to-end framework named Recurrent Saliency Transformation Network (RSTN) for seg- menting tiny and/or variable targets. RSTN is a coarse-to-fine approach, which uses prediction from the first (coarse) stage to shrink the input region for the second (fine) stage. A saliency transformation module is inserted between these two stages, so that (i) the coarse-scaled segmentation mask can be transferred as spatial weights and applied to the fine stage; and (ii) the gradients can be back-propagated from the loss layer to the entire network, so that the two stages are optimized in a joint manner. In the testing stage, we perform segmentation iteratively to improve accuracy.
    Recurrent Saliency Transformation Network for Tiny Target Segmentation in Abdominal CT Scans
    Lingxi Xie, Qihang Yu, Yan Wang, Yuyin Zhou, Elliot K. Fishman, and Alan L. Yuille
    IEEE Trans Med Imaging. 2019 Jul 23. doi: 10.1109/TMI.2019.2930679
  • “In this extended journal paper, we allow a gradual optimization to improve the stability of RSTN, and introduce a hierarchical version named H-RSTN to segment tiny and variable neoplasms such as pancreatic cysts. Experiments are performed on several CT datasets, including a public pancreas segmentation dataset, our own multi-organ dataset, and a cystic pancreas dataset. In all these cases, RSTN outperforms the baseline (a stage-wise coarse-to-fine approach) significantly. Confirmed by the radiologists in our team, these promising segmentation results can help early diagnosis of pancreatic cancer.”
    Recurrent Saliency Transformation Network for Tiny Target Segmentation in Abdominal CT Scans
    Lingxi Xie, Qihang Yu, Yan Wang, Yuyin Zhou, Elliot K. Fishman, and Alan L. Yuille
    IEEE Trans Med Imaging. 2019 Jul 23. doi: 10.1109/TMI.2019.2930679
  • “Motivated by the above, we propose a Recurrent Saliency Transformation Network (RSTN) for segmenting very small targets. The chief innovation lies in the mechanism to relate the coarse and fine stages with a saliency transformation module, which repeatedly transforms the segmentation probability map as spatial weights, from the previous iterations to the current iteration. In the training process, the differentiability of this module makes it possible to optimize the coarse-scaled and fine-scaled networks in a joint manner, so that the overall mod- el gets improved after being aware of a global optimization goal.”
    Recurrent Saliency Transformation Network for Tiny Target Segmentation in Abdominal CT Scans
    Lingxi Xie, Qihang Yu, Yan Wang, Yuyin Zhou, Elliot K. Fishman, and Alan L. Yuille
    IEEE Trans Med Imaging. 2019 Jul 23. doi: 10.1109/TMI.2019.2930679

  • Recurrent Saliency Transformation Network for Tiny Target Segmentation in Abdominal CT Scans
    Lingxi Xie, Qihang Yu, Yan Wang, Yuyin Zhou, Elliot K. Fishman, and Alan L. Yuille
    IEEE Trans Med Imaging. 2019 Jul 23. doi: 10.1109/TMI.2019.2930679

  • Recurrent Saliency Transformation Network for Tiny Target Segmentation in Abdominal CT Scans
    Lingxi Xie, Qihang Yu, Yan Wang, Yuyin Zhou, Elliot K. Fishman, and Alan L. Yuille
    IEEE Trans Med Imaging. 2019 Jul 23. doi: 10.1109/TMI.2019.2930679

  • Recurrent Saliency Transformation Network for Tiny Target Segmentation in Abdominal CT Scans
    Lingxi Xie, Qihang Yu, Yan Wang, Yuyin Zhou, Elliot K. Fishman, and Alan L. Yuille
    IEEE Trans Med Imaging. 2019 Jul 23. doi: 10.1109/TMI.2019.2930679
  • “We present the Recurrent Saliency Transformation Network, which enjoys three advantages. (i) Benefited by a (recurrent) global energy function, it is easier to generalize our models from training data to testing data. (ii) With joint optimization over two networks, both of them get improved individually. (iii) By incorporating multi-stage visual cues, more accurate segmentation results are obtained.”
    Recurrent Saliency Transformation Network for Tiny Target Segmentation in Abdominal CT Scans
    Lingxi Xie, Qihang Yu, Yan Wang, Yuyin Zhou, Elliot K. Fishman, and Alan L. Yuille
    IEEE Trans Med Imaging. 2019 Jul 23. doi: 10.1109/TMI.2019.2930679
  • ”We aim at segmenting a wide variety of organs, including tiny targets (e.g., adrenal gland) and neoplasms (e.g., pancreatic cyst), from abdominal CT scans. This is a challenging task in two aspects. First, some organs (e.g., the pancreas), are highly variable in both anatomy and geometry, and thus very difficult to depict. Second, the neoplasms often vary a lot in its size, shape, as well as its location within the organ. Third, the targets (organs and neoplasms) can be considerably small compared to the human body, and so standard deep networks for segmentation are often less sensitive to these targets and thus predict less accurately especially around their boundaries.”
    Recurrent Saliency Transformation Network for Tiny Target Segmentation in Abdominal CT Scans
    Lingxi Xi, Qihang Yu, Yan Wang, Yuyin Zhou, Elliot K. Fishman, and Alan L. Yuille
    IEEE TRANSACTIONS ON MEDICAL IMAGING (in press)
  • “In conclusion, our study provided preliminary evidence that textural features derived from CT images were useful in differential diagnosis of pancreatic mucinous cystadenomas and serous cystadenomas, which may provide a non-invasive approach to determine whether surgery is needed in clinical practice. However, multicentre studies with larger sample size are needed to confirm these results.”
    Discrimination of Pancreatic Serous Cystadenomas From Mucinous Cystadenomas With CT Textural Features: Based on Machine Learning
    Yang J et al.
    Front. Oncol., 12 June 2019 /doi.org/10.3389/fonc.2019.00494
  • Results: Only 31 of 102 serous cystic neoplasm cases in this study were recognized correctly by clinicians before the surgery. Twenty-two features were selected from the radiomics system after 100 bootstrapping repetitions of the least absolute shrinkage selection operator regression. The diagnostic scheme performed accurately and robustly, showing the area under the receiver operating characteristic curve 1⁄4 0.767, sensitivity 1⁄4 0.686, and specificity 1⁄4 0.709. In the independent validation cohort, we acquired similar results with receiver operating characteristic curve 1⁄4 0.837, sensitivity 1⁄4 0.667, and specificity 1⁄4 0.818.
    Conclusion: The proposed radiomics-based computer-aided diagnosis scheme could increase preoperative diagnostic accuracy and assist clinicians in making accurate management decisions.
    Computer-Aided Diagnosis of Pancreas Serous Cystic Neoplasms: A Radiomics Method on Preoperative MDCT Images
    Ran Wei et al.
    Technology in Cancer Research & Treatment
    Volume 18: 1-9; 2019
  • “A total of 17 intensity and texture features were selected, showing difference between SCNs and non-SCNs. Typically, the intensity T-range, wavelet intensity T-median, and wavelet neighborhood gray-tone difference matrix (NGTDM) busyness were the most distinguishable.”
    Computer-Aided Diagnosis of Pancreas Serous Cystic Neoplasms: A Radiomics Method on Preoperative MDCT Images
    Ran Wei et al.
    Technology in Cancer Research & Treatment
    Volume 18: 1-9; 2019
  • “In our retrospective study of 260 patients with PCN, we were surprised to find that the overall preoperative diagnostic accuracy by clinicians was 37.3% (97 of 260), and only 30.4% (31 of 102) of SCN cases were correctly diagnosed. This meant that more than two-thirds of patients with SCN suffered unnecessary pancreatic resection.”
    Computer-Aided Diagnosis of Pancreas Serous Cystic Neoplasms: A Radiomics Method on Preoperative MDCT Images
    Ran Wei et al.
    Technology in Cancer Research & Treatment
    Volume 18: 1-9; 2019
  • “Furthermore, radiomics high-throughput features containing intensity features, texture features, and their wavelet decomposition forms fully utilized image information and obtained more image details that were hard to discover with the naked human eyes.”
    Computer-Aided Diagnosis of Pancreas Serous Cystic Neoplasms: A Radiomics Method on Preoperative MDCT Images
    Ran Wei et al.
    Technology in Cancer Research & Treatment
    Volume 18: 1-9; 2019
  • “In conclusion, our study proposed a radiomics-based CAD scheme and stressed the role of radiomics analysis as a novel noninvasive method for improving the preoperative diagnostic accuracy of SCNs. In all, 409 quantitative features were auto- matically extracted, and a feature subset containing the 22 most statistically significant features was selected after 100 boot- strapping repetitions. Our proposed method improved the diag- nostic accuracy and performed well in all metrics, with AUC of 0.767 in the cross-validation cohort and 0.837 in the independent validation cohort. This demonstrated that our CAD scheme could provide a powerful reference for the diagnosis of clinicians to reduce misjudgment and avoid overtreatment.”
    Computer-Aided Diagnosis of Pancreas Serous Cystic Neoplasms: A Radiomics Method on Preoperative MDCT Images
    Ran Wei et al.
    Technology in Cancer Research & Treatment
    Volume 18: 1-9; 2019
  • “In conclusion, our study proposed a radiomics-based CAD scheme and stressed the role of radiomics analysis as a novel noninvasive method for improving the preoperative diagnostic accuracy of SCNs. In all, 409 quantitative features were auto- matically extracted, and a feature subset containing the 22 most statistically significant features was selected after 100 boot- strapping repetitions. Our proposed method improved the diag- nostic accuracy and performed well in all metrics, with AUC of 0.767 in the cross-validation cohort and 0.837 in the independent validation cohort. This demonstrated that our CAD scheme could provide a powerful reference for the diagnosis of clinicians to reduce misjudgment and avoid overtreatment.”
    Computer-Aided Diagnosis of Pancreas Serous Cystic Neoplasms: A Radiomics Method on Preoperative MDCT Images
    Ran Wei et al.
    Technology in Cancer Research & Treatment
    Volume 18: 1-9; 2019
  • “In this paper, we adopt 3D CNNs to segment the pancreas in CT images. Although deep neural networks have been proven to be very effective on many 2D vision tasks, it is still challenging to apply them to 3D applications due to the limited amount of annotated 3D data and limited computational resources. We propose a novel 3D-based coarse- to-fine framework for volumetric pancreas segmentation to tackle these challenges. The proposed 3D-based framework outperforms the 2D counterpart to a large margin since it can leverage the rich spatial information along all three axes.”


    A 3D Coarse-to-Fine Framework for Automatic Pancreas Segmentation 
Zhuotun Zhu, Yingda Xia, Wei Shen, Elliot K. Fishman, Alan L. Yuille
arXiv:1712.00201v1 [cs.CV] 1 Dec 2017 

  • “In this work, we proposed a novel 3D network called “ResDSN” integrated with a coarse-to-fine framework to simultaneously achieve high segmentation accuracy and low time cost. The backbone network “ResDSN” is carefully designed to only have long residual connections for efficient inference. To our best knowledge, we are the first to segment the challenging pancreas using 3D networks which leverage the rich spatial information to achieve the state-of- the-art.”

    
A 3D Coarse-to-Fine Framework for Automatic Pancreas Segmentation 
Zhuotun Zhu, Yingda Xia, Wei Shen, Elliot K. Fishman, Alan L. Yuille
arXiv:1712.00201v1 [cs.CV] 1 Dec 2017 

  • “To address these issues, we propose a concise and effective framework based on 3D deep networks for pancreas segmentation, which can simultaneously achieve high seg- mentation accuracy and low time cost. Our framework is formulated in a coarse-to-fine manner. In the training stage, we first train a 3D FCN from the sub-volumes sampled from an entire CT volume. We call this ResDSN Coarse model, which aims to obtain the rough location of the target pancreas from the whole CT volume by making full use of the overall 3D context. Then, we train another 3D FCN from the sub-volumes sampled only from the ground truth bound- ing boxes of the target pancreas. We call this the ResDSN Fine model, which can refine the segmentation based on the coarse result.”


    A 3D Coarse-to-Fine Framework for Automatic Pancreas Segmentation 
Zhuotun Zhu, Yingda Xia, Wei Shen, Elliot K. Fishman, Alan L. Yuille
arXiv:1712.00201v1 [cs.CV] 1 Dec 2017 

  • “This work is motivated by the difficulty of small organ segmentation. As the target is often small, it is required to 
focus on a local input region, but sometimes the network is confused due to the lack of contextual information. We present the Recurrent Saliency Transformation Network, which enjoys three advantages. (i) Benefited by a (recurrent) global energy function, it is easier to generalize our models from training data to testing data. (ii) With joint optimization over two networks, both of them get improved individually. (iii) By incorporating multi-stage visual cues, more accurate segmentation results are obtained. As the fine stage is less likely to be confused by the lack of contexts, we also observe better convergence during iterations.”


    Recurrent Saliency Transformation Network: Incorporating Multi-Stage Visual Cues for Small Organ Segmentation 
Qihang Yu, Lingxi Xie, Yan Wang, Yuyin Zhou, Elliot K. Fishman, Alan L. Yuille
arXiv:1709.04518v3 [cs.CV] 18 Nov 2017
  • “This paper presents a Recurrent Saliency Transforma- tion Network. The key innovation is a saliency transfor- mation module, which repeatedly converts the segmentation probability map from the previous iteration as spatial weights and applies these weights to the current iteration. This brings us two-fold benefits. In training, it allows joint optimization over the deep networks dealing with different input scales. In testing, it propagates multi-stage visual information throughout iterations to improve segmentation accuracy.”


    Recurrent Saliency Transformation Network: Incorporating Multi-Stage Visual Cues for Small Organ Segmentation 
Qihang Yu, Lingxi Xie, Yan Wang, Yuyin Zhou, Elliot K. Fishman, Alan L. Yuille
arXiv:1709.04518v3 [cs.CV] 18 Nov 2017
  • “Automatic segmentation of an organ and its cystic region is a prerequisite of computer-aided diagnosis. In this paper, we focus on pancreatic cyst segmentation in abdominal CT scan. This task is important and very useful in clinical practice yet challenging due to the low contrast in boundary, the variability in location, shape and the different stages of the pancreatic cancer. Inspired by the high relevance between the location of a pancreas and its cystic region, we introduce extra deep supervision into the segmentation network, so that cyst segmentation can be improved with the help of relatively easier pancreas segmentation.”


    Deep Supervision for Pancreatic Cyst Segmentation in Abdominal CT Scans 
Yuyin Zhou, Lingxi Xie, Elliot K. Fishman, and Alan L. Yuille 
(in) Medical Image Computing and Computer Assisted Intervention − MICCAI 2017
page 222-231
  • “This paper presents the first system for pancreatic cyst segmentation which can work without human assistance on the testing stage. Motivated by the high relevance of a cystic pancreas and a pancreatic cyst, we formulate pancreas segmentation as an explicit variable in the formulation, and introduce deep supervision to assist the network training process. The joint optimization can be factorized into two stages, making our approach very easy to implement. We collect a dataset with 131 pathological cases. Based on a coarse-to-fine segmentation algorithm, our approach produces reasonable cyst segmentation results. It is worth emphasizing that our approach does not require any extra human annotations on the testing stage, which is especially practical in assisting common patients in cheap and periodic clinical applications.”

    
Deep Supervision for Pancreatic Cyst Segmentation in Abdominal CT Scans 
Yuyin Zhou, Lingxi Xie, Elliot K. Fishman, and Alan L. Yuille 
(in) Medical Image Computing and Computer Assisted Intervention − MICCAI 2017
page 222-231
  • “The pancreas is a highly deformable organ that has a shape and location that is greatly influenced by the presence of adjacent struc- tures. This makes automated image analysis of the pancreas extremely challenging. A number of different approaches have been taken to automated pancreas analysis, in- cluding the use of anatomic atlases, the loca- tion of the splenic and portal veins, and state- of-the-art computer science methods such as deep learning.”

    Progress in Fully Automated Abdominal CT Interpretation
Summers RM
AJR 2016; 207:67–79
  • “A recent advance in computer science is the refinement of neural networks, a type of machine learning classifier used to make decisions from data. This refine- ment, known generically as deep learn- ing but more specifically as convolutional neural networks, has shown dramatic improvements in automated intelligence applications. Initially drawing attention for impressive improvements in speech recognition and natural image interpretation, deep learning is now being applied to medical images, as described already in the sections on the pancreas and colitis.” 


    Progress in Fully Automated Abdominal CT Interpretation
Summers RM
AJR 2016; 207:67–79

Privacy Policy

© 1999-2022 Elliot K. Fishman, MD, FACR. All rights reserved.