Rare Disease Classification in Video Capsule Endoscopy
Abstract
Video capsule endoscopy offers a non-invasive way to examine the small bowel, however, the rare but clinically important findings that appear only in less than 1\% of labelled frames, provide very less signal for detection. As a result, while models can achieve high overall performance, they almost completely miss these rare findings. We condition each frame's decision on evidence from the rest of that patient's video, not on the frame's own posterior alone, supplying information a single frozen frame can never contain. Building upon Expected Gain Arbitration, a per-frame, macro-F1-optimal correction over the confusion matrix, we further compose it with video-disease-presence conditional resulting in our decoding method Hierarchical Conditional-Expected-gain (HCE). We evaluate HCE on two independent VCE datasets, Kvasir-Capsule and GALAR, against a static temporal-smoothing baseline and an optimal-transport (OT) frame-level correction. HCE significantly improves rare-class recall over a temporal-smoothing baseline on GALAR (2.7--3.1x; 95\% CI excludes zero in both scopes) and shows a similar, though not statistically confirmed, gain on Kvasir-Capsule (up to 3.4x). Against OT, HCE still improves rare-class recall in all four evaluation settings (+0.016 to +0.029), and additionally improves macro-F1 and MCC.
Related
- Artificial Intelligence Versus Conventional Scoring for Assessing Small Bowel Capsule Endoscopy Cleanliness in Crohn's Disease
- Concordancia diagnóstica entre la cápsula endoscópica y la enteroscopía de doble balón en la hemorragia digestiva de intestino medio en un hospital de referencia en Perú Diagnostic Agreement Between Capsule Endoscopy and Double-Balloon Enteroscopy in Mid-Gut Gastrointestinal Bleeding at a Referral Hospital in Peru
- Surgical Success in Obstructive Sleep Apnea Patients Using Preoperative Drug‐Induced Sleep Endoscopy and Awake Endoscopy Considering Disease Severity
- A liquid- and capsule-compatible hydroxyurea electronic monitor to identify non-adherent youth with sickle cell disease
- Lens Capsule microRNA Profiling Across the Primary Angle-Closure Disease Spectrum: Association with Anterior Segment Crowding
BibTeX
@misc{patel2026rare,
title = {{Rare Disease Classification in Video Capsule Endoscopy}},
author = {Amit Kumar Patel and Debesh Jha},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.03733},
eprint = {2610.03733},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.03733},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Rare Disease Classification in Video Capsule Endoscopy
AU - Amit Kumar Patel
AU - Debesh Jha
PY - 2026
DA - 2026/10/01
PB - arXiv
DO - 10.48550/arxiv.2610.03733
UR - https://arxiv.org/abs/2610.03733
AB - Video capsule endoscopy offers a non-invasive way to examine the small bowel, however, the rare but clinically important findings that appear only in less than 1\% of labelled frames, provide very less signal for detection. As a result, while models can achieve high overall performance, they almost completely miss these rare findings. We condition each frame's decision on evidence from the rest of that patient's video, not on the frame's own posterior alone, supplying information a single frozen frame can never contain. Building upon Expected Gain Arbitration, a per-frame, macro-F1-optimal correction over the confusion matrix, we further compose it with video-disease-presence conditional resulting in our decoding method Hierarchical Conditional-Expected-gain (HCE). We evaluate HCE on two independent VCE datasets, Kvasir-Capsule and GALAR, against a static temporal-smoothing baseline and an optimal-transport (OT) frame-level correction. HCE significantly improves rare-class recall over a temporal-smoothing baseline on GALAR (2.7--3.1x; 95\% CI excludes zero in both scopes) and shows a similar, though not statistically confirmed, gain on Kvasir-Capsule (up to 3.4x). Against OT, HCE still improves rare-class recall in all four evaluation settings (+0.016 to +0.029), and additionally improves macro-F1 and MCC.
ER -