Conditional Conformal Risk Adaptation

2025-04-10

AI Summary

To address the imbalance in image-level conditional risks (e.g., false negative rate) and the difficulty of satisfying clinical personalized decision-making requirements in medical image segmentation, this paper proposes an adaptive conformal risk control framework. We first establish a unified theoretical connection between weighted quantiles and conformal prediction. Then, we design two novel methods: CCRA, enhanced with probability calibration for pixel-wise calibration, and CCRA-S, incorporating hierarchical image feature representation to optimize stratified risk thresholds. Evaluated on polyp segmentation, our approach strictly guarantees marginal risk constraints while reducing inter-image false negative rate variance by over 40%, significantly outperforming existing conformal methods. This work introduces a verifiable and controllable uncertainty quantification paradigm for high-stakes medical AI systems, enabling rigorous, clinically adaptable risk management.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Segmentation
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Uncertainty quantification is becoming increasingly important in image segmentation, especially for high-stakes applications like medical imaging. While conformal risk control generalizes conformal prediction beyond standard miscoverage to handle various loss functions such as false negative rate, its application to segmentation often yields inadequate conditional risk control: some images experience very high false negative rates while others have negligibly small ones. We develop Conformal Risk Adaptation (CRA), which introduces a new score function for creating adaptive prediction sets that significantly improve conditional risk control for segmentation tasks. We establish a novel theoretical framework that demonstrates a fundamental connection between conformal risk control and conformal prediction through a weighted quantile approach, applicable to any score function. To address the challenge of poorly calibrated probabilities in segmentation models, we introduce a specialized probability calibration framework that enhances the reliability of pixel-wise inclusion estimates. Using these calibrated probabilities, we propose Calibrated Conformal Risk Adaptation (CCRA) and a stratified variant (CCRA-S) that partitions images based on their characteristics and applies group-specific thresholds to further enhance conditional risk control. Our experiments on polyp segmentation demonstrate that all three methods (CRA, CCRA, and CCRA-S) provide valid marginal risk control and deliver more consistent conditional risk control across diverse images compared to standard approaches, offering a principled approach to uncertainty quantification that is particularly valuable for high-stakes and personalized segmentation applications.

Problem

• Improves conditional risk control in image segmentation
• Addresses poorly calibrated probabilities in segmentation models
• Enhances uncertainty quantification for high-stakes medical imaging

Innovation

• Adaptive prediction sets improve risk control
• Weighted quantile links risk control and prediction
• Calibrated probabilities enhance pixel-wise reliability