Beyond Uncertainty: Generalizable Failure Monitoring for Surgical Segmentation under Acquisition Degradation
2026-08-17 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors show that surgical image segmentation models can make wrong predictions even when they seem very sure. They introduce TCSR-Monitor, a system that checks for mistakes by looking at shape, timing, and image quality cues in addition to confidence, without needing to change the original model or know the true answer. Their tests show it works better than just using confidence alone, but it can still falsely warn on some good images and has limits when applied to different models. They highlight that using multiple signals helps detect failures, but more improvements are needed to reduce false alarms and improve model transferability.
surgical segmentationmodel confidencefailure detectiondistribution shiftconformal calibrationimage corruptiontemporal consistencypost-hoc monitoringfalse alarmszero-shot transfer
Authors
Hieu D. Pham, Dang P. M. Cao, Thanh Trung Huynh
Abstract
Surgical segmentation networks can fail silently under acquisition degradation: predicted masks may be wrong even when model confidence remains high. Existing deployment-time monitors rely primarily on uncertainty estimates and can therefore miss confident failures. We present TCSR-Monitor (Temporal Conformal Surgical Risk Monitor), a post-hoc failure-monitoring framework that combines confidence with observable shape, temporal-consistency, and image-quality cues. TCSR-Monitor wraps a frozen segmentation model, requires no model internals, and operates without ground truth at deployment. We also introduce a validation protocol to assess whether alarms remain credible under distribution shift. On EndoVis 2017, leave-one-corruption-out evaluation shows that TCSR-Monitor generalizes to unseen acquisition degradations and substantially outperforms confidence-based baselines. A circularity control confirms that it predicts segmentation failure rather than simply detecting corrupted images. Mondrian conformal calibration balances miss-rates across degradation severities, but a single global threshold still produces false alarms on up to 40% of correctly segmented frames at moderate corruption. Zero-shot transfer to SAM2 demonstrates feature portability, although entropy outperforms the transferred monitor at both evaluated thresholds. Overall, reliable monitoring under acquisition degradation benefits from complementary observable signals beyond confidence alone, but substantial false-alarm and transfer limitations remain.