Clinician input guides safer medical image segmentation under uncertainty
From Few-Shot Segmentation to Clinician-in-the-Loop Medical Image Analysis
Computer Vision and Pattern Recognition
Summary
Medical image analysis often tries to identify regions in images with very limited examples, but this can be tricky when images are unclear or different from the training data. The authors rethink this as a step-by-step process where the computer asks a doctor for help only when it is likely to improve the results safely. Their approach considers that doctors' answers are helpful but can sometimes be wrong, and it balances expert time with the risks of mistakes. This framework helps computers and clinicians work together more efficiently to handle tough medical images.
What this means in practice
- •For radiology teams: Integrate expert feedback selectively into imaging software when it can reduce diagnostic risks in ambiguous or poor-quality scans.
- •For medical imaging software developers: Build adaptive segmentation tools that request clinician input strategically to maintain safety across different hospitals and scan types.
A position paper. It proposes an approach and reports no results.
Authors
Yazhou Zhu
Abstract
Few-shot medical image segmentation (FSMIS) seeks to delineate unseen structures from a small support set, but its standard formulation fixes task-defining evidence before inference. This assumption is fragile when query cases exhibit acquisition shift, atypical pathology, ambiguous boundaries, or poor image quality. Prototype learning, cross-domain matching, interactive segmentation, uncertainty estimation, test-time adaptation, and promptable foundation models address parts of this problem, yet have not been jointly evaluated under a common model of expert attention and clinical risk. This Perspective reframes FSMIS as a sequential clinician-model decision problem with a static support budget $K$ and a distinct interaction budget $B$. At each step, a system accepts the current segmentation, requests feedback, or defers to full expert review. Queries vary in location and modality and are selected by response-conditioned net expected value of information; clinician-provided feedback informs bounded adaptation only after prespecified provenance, consistency, and safety gates. The framework separates distributional atypicality from predicted clinical failure and treats clinician responses as informative but fallible observations. We synthesize the transition from few-shot and cross-domain segmentation to interactive and selective adaptation, delineate the integration gap, and define four research directions with falsifiable hypotheses. Evaluation spans external-domain calibration, quality-effort trade-offs, reader studies, and prospective workflow assessment. The central claim is not that interaction alone resolves domain shift, but that scarce expert attention should be allocated only when it is expected to reduce clinically relevant risk.