To develop and preliminarily validate Eval-Model, a deep-learning framework for automated Lewis score estimation from small-bowel capsule endoscopy in Crohn disease.
Approach:
Model: Eval-Model combines lesion detection and localization, pixelwise segmentation of edema and ulcers, stenosis classification, and LSTM-based temporal regression to estimate an examination-level Lewis score.
Architecture: The framework uses a customized YOLO-style detection backbone with multiscale outputs and integrates lesion and stenosis features across sequential capsule endoscopy frames.
Interpretation:
The provided context describes the study rationale and model architecture but does not report cohort details, validation results, or performance findings.
Limitations:
The supplied article context ends during the Methods section and does not include study results or the authors’ stated limitations.
Pancreatic adenocarcinoma, the most common form of pancreatic cancer, accounts for roughly 90 to 95 percent of cases. It's notoriously difficult to catch early, because symptoms often don't appear until the disease has already spread
