Imaging Pearls ❯ Stomach ❯ AI and the Stomach
|
-- OR -- |
|
- Purpose: To develop and validate a deep learning–based approach, Gastric Neoplasm Detection with Artificial Intelligence (GANDA), for automated detection, diagnosis, and segmentation of gastric neoplasms at clinical routine contrast-enhanced CT
Results: A total of 4606 patients were included in the study (median age, 57 years [IQR: 48–66]; 2554 male patients). In the internal test cohort (n = 266), GANDA achieved 87.3% sensitivity and 87.2% specificity for tumor detection. The model demonstrated significantly higher diagnostic accuracy (top-1 accuracy, 85.3%; 95% CI: 81.2, 89.1) compared with radiologists (mean accuracy, 74.2%; 95% CI: 70.5, 77.6; P = .002). In the external test cohort (n = 2657), GANDA distinguished between patients with gastric neoplasms and controls with 77.4% sensitivity and 89.8% specificity. The mean Dice coefficient in the internal test cohort was 0.52 for gastric cancer and 0.45 for non–gastric cancer. In the real-world test cohort (n = 7695), GANDA achieved 83.2% sensitivity and 93.1% specificity for tumor detection.
Conclusion: GANDA enabled the detection and segmentation of gastric neoplasms at routine clinical CT scans.
Gastric Neoplasm Detection at Contrast-enhanced CT with Deep Learning
Xin Chen, MD • Yingda Xia • Lisha Yao ett al.
Radiology: Artificial Intelligence 2026; 8(1):e250145 - DL-based approaches offer a paradigm-shifting solution to harness the untapped potential of routine abdominal CT for gastric lesion detection. By prioritizing suspicious findings and providing quantitative malignancy risk stratification, such systems may serve as vigilant “second readers,” particularly valuable for subtle lesions overlooked at initial interpretation. Despite these prospects, no comprehensive AI framework currently exists for systematic gastric lesion detection and characterization at routine venous-phase CT. Therefore, in this study, we developed and validated, to our knowledge, the first DL-based approach, Gastric Neoplasm Detection with Artificial Intelligence (GANDA), to detect, diagnose, and segment gastric neoplasms (gastric cancer and non–gastric cancer) on contrast-enhanced CT images.
Gastric Neoplasm Detection at Contrast-enhanced CT with Deep Learning
Xin Chen, MD • Yingda Xia • Lisha Yao ett al.
Radiology: Artificial Intelligence 2026; 8(1):e250145 - ■ Gastric Neoplasm Detection with Artificial Intelligence (GANDA) was developed using contrast-enhanced CT images from 798 patients with gastric neoplasms and 885 controls.
■ On the external test dataset with 2657 CT studies (1300 patients and 1357 controls), GANDA achieved 77.4% sensitivity and 89.8% specificity for neoplasm detection.
■ On a real-world test dataset of 7695 consecutive CT scans, GANDA achieved 83.2% sensitivity and 93.1% specificity for neoplasm detection and 85.7% sensitivity for gastric cancer diagnosis.
Gastric Neoplasm Detection at Contrast-enhanced CT with Deep Learning
Xin Chen, MD • Yingda Xia • Lisha Yao ett al.
Radiology: Artificial Intelligence 2026; 8(1):e250145 - Consistent with clinical scenarios, the false-positive cases were primarily attributed to suboptimal gastric distention, high-attenuation gastric contents, and confusion with adjacent organs. The main causes of false-negative cases were small-sized lesions and lesions partially obscured by a uniformly thickened gastric wall, often due to suboptimal gastric distention . Although associated with false positives and false negatives, GANDA contributed to reducing missed diagnoses by correctly detecting approximately 80% of gastric tumors and directing radiologists’ attention to them.
Gastric Neoplasm Detection at Contrast-enhanced CT with Deep Learning
Xin Chen, MD • Yingda Xia • Lisha Yao ett al.
Radiology: Artificial Intelligence 2026; 8(1):e250145 - Our study had a few limitations. First, the retrospective design introduces potential selection bias. Temporal heterogeneity in imaging data is a particular concern. However, the proven longitudinal consistency of CT hardware (>10-year service life) with fixed acquisition settings likely mitigated substantial technical variations. Second, early-stage gastric cancer and non–gastric cancer detection performance was suboptimal compared with gastric cancer, attributable to the smaller training cohort. Third, the reader study utilized a neoplasm-enriched dataset with higher tumor prevalence than real-world settings. This deliberate enrichment aligns with established AI validation standards, providing preliminary evidence for AI-assistedlesion detection improvement.
Gastric Neoplasm Detection at Contrast-enhanced CT with Deep Learning
Xin Chen, MD • Yingda Xia • Lisha Yao ett al.
Radiology: Artificial Intelligence 2026; 8(1):e250145 - In conclusion, we demonstrated that the AI model trained on routine CT images from multiple centers could achieve high accuracy in distinguishing controls, gastric cancer, and non–gastric cancer. The newly developed AI model, GANDA, exhibited an excellent diagnostic performance comparable or better to that of expert radiologists. In future work, we will enhance GANDA’s sensitivity for early-stage gastric cancer and non–gastric cancer by augmenting the training dataset with additional early-stage and non–gastric cancer cases. Concurrently, based on our observation that optimal gastric distention improves early gastric cancer detection sensitivity, our prospective trial protocol will standardize pre-CT gastric distention.
Gastric Neoplasm Detection at Contrast-enhanced CT with Deep Learning
Xin Chen, MD • Yingda Xia • Lisha Yao ett al.
Radiology: Artificial Intelligence 2026; 8(1):e250145
- Background: Image recognition using artificial intelligence with deep learning through convolutional neural networks (CNNs) has dramatically improved and been increasingly applied to medical fields for diagnostic imaging. We developed a CNN that can automatically detect gastric cancer in endoscopic images.
Methods: A CNN-based diagnostic system was constructed based on Single Shot MultiBox Detector architecture and trained using 13,584 endoscopic images of gastric cancer. To evaluate the diagnostic accuracy, an independent test set of 2296 stomach images collected from 69 consecutive patients with 77 gastric cancer lesions was applied to the constructed CNN.
Application of artificial intelligence using a convolutional neural network for detecting gastric cancer in endoscopic images
Toshiaki Hirasawa et al.
Gastric Cancer (2018) 21:653–660 - Results: The CNN required 47 s to analyze 2296 test images. The CNN correctly diagnosed 71 of 77 gastric cancer lesions with an overall sensitivity of 92.2%, and 161 non-cancerous lesions were detected as gastric cancer, resulting in a positive predictive value of 30.6%. Seventy of the 71 lesions (98.6%) with a diameter of 6 mm or more as well as all invasive cancers were correctly detected. All missed lesions were superficially depressed and differentiated-type intramucosal cancers that were difficult to distinguish from gastritis even for experienced endoscopists. Nearly half of the false-positive lesions were gastritis with changes in color tone or an irregular mucosal surface.
Conclusion: The constructed CNN system for detecting gastric cancer could process numerous stored endoscopic images in a very short time with a clinically relevant diagnostic ability. It may be well applicable to daily clinical practice to reduce the burden of endoscopists.
Application of artificial intelligence using a convolutional neural network for detecting gastric cancer in endoscopic images
Toshiaki Hirasawa et al.
Gastric Cancer (2018) 21:653–660 - “In conclusion, we developed a CNN system for detecting gastric cancer using stored endoscopic images, which processed extensive independent images in a very short time. The clinically relevant diagnostic ability of the CNN offers a promising applicability to daily clinical practice for reducing the burden of endoscopists as well as telemedicine in remote and rural areas as well as in developing countries where the number of endoscopists is limited.”
Application of artificial intelligence using a convolutional neural network for detecting gastric cancer in endoscopic images
Toshiaki Hirasawa et al.
Gastric Cancer (2018) 21:653–660