Anatomically-aware conformal prediction for medical image segmentation with random walks

TL;DR

This paper introduces RW-CP, a novel conformal prediction framework that enforces spatial coherence in medical image segmentation, improving anatomical validity and segmentation quality while maintaining statistical guarantees.

Contribution

It proposes a model-agnostic random walk-based conformal prediction method that enhances spatial coherence and stability in uncertainty quantification for medical image segmentation.

Findings

Improves segmentation quality by up to 35.4% over standard methods.

Maintains rigorous marginal coverage with enhanced anatomical coherence.

Reduces sensitivity to calibration parameters, ensuring stable predictions.

Abstract

The reliable deployment of deep learning in medical imaging requires uncertainty quantification that provides rigorous error guarantees while remaining anatomically meaningful. Conformal prediction (CP) is a powerful distribution-free framework for constructing statistically valid prediction intervals. However, standard applications in segmentation often ignore anatomical context, resulting in fragmented, spatially incoherent, and over-segmented prediction sets that limit clinical utility. To bridge this gap, this paper proposes Random-Walk Conformal Prediction (RW-CP), a model-agnostic framework which can be added on top of any segmentation method. RW-CP enforces spatial coherence to generate anatomically valid sets. Our method constructs a k-nearest neighbour graph from pre-trained vision foundation model features and applies a random walk to diffuse uncertainty. The random walk diffusion regularizes the non-conformity scores, making the prediction sets less sensitive to the conformal calibration parameter $\lambda$, ensuring more stable and continuous anatomical boundaries. RW-CP maintains rigorous marginal coverage while significantly improving segmentation quality. Evaluations on multi-modal public datasets show improvements of up to $35.4\%$ compared to standard CP baselines, given an allowable error rate of $\alpha=0.1$.