Quantification of Uncertainty with Adversarial Models in Medical Image Segmentation

2026-06-17

AI Summary

Standard deep learning models for medical image segmentation often suffer from poor calibration, leading to overconfident predictions and unreliable pixel-level uncertainty quantification—particularly at pathological boundaries where true risks are frequently overlooked. To address this, this work proposes the QUAM-SM framework, which uniquely introduces adversarial perturbations during the post-processing stage. By performing targeted adversarial searches, the method identifies “adversarially vulnerable” pixels, effectively disentangling and precisely localizing both epistemic and aleatoric uncertainties. Validated against multi-expert annotations, QUAM-SM demonstrates significant improvements over existing approaches on two public datasets, markedly enhancing the reliability of uncertainty estimation and sensitivity to pathological boundaries.

Abstract

Reliable pixel-level uncertainty quantification holds the potential to transform clinical workflows by enabling high-fidelity longitudinal monitoring and distinguishing true pathological changes from artifacts. Ideally, these models provide the stability required for critical treatment planning and surgical intervention. However, standard deep learning models often suffer from miscalibration, yielding overconfident predictions that mask underlying vulnerabilities at subtle pathological boundaries. To address this, we propose QUAM-SM, a post-hoc framework using targeted adversarial search to identify "adversarially fragile" pixels. By actively seeking perturbations that expose predictive instability, our method highlights regions where decisions are most vulnerable to being flipped. Importantly, the framework disentangles epistemic uncertainty from aleatoric uncertainty. Experiments on two public datasets with multiple expert annotations demonstrate that QUAM-SM outperforms both standard and recent uncertainty estimation approaches in terms of reliability and boundary sensitivity. Code is available at https://github.com/HanaJebril/quam_sm

Problem

• uncertainty quantification
• medical image segmentation
• adversarial models
• miscalibration
• pixel-level uncertainty

Innovation

• adversarial uncertainty quantification
• medical image segmentation
• epistemic vs aleatoric uncertainty
• post-hoc calibration
• boundary-sensitive uncertainty