Generalizing Abstention for Noise-Robust Learning in Medical Image Segmentation

2026-01-20

AI Summary

Label noise in medical image segmentation often leads to model overfitting and degraded generalization performance. To address this issue, this work proposes a general and modular abstention learning framework that systematically guides the model to disregard corrupted samples during training. The framework incorporates an abstention mechanism regularized by an information-theoretic penalty and a power-law-based adaptive algorithm for tuning the abstention penalty. Building upon this framework, three novel noise-robust loss functions—GAC, SAC, and ADS—are developed and seamlessly integrated into mainstream segmentation architectures. Experimental results on the CaDIS and DSAD datasets demonstrate that the proposed approach significantly outperforms existing non-abstention baselines, particularly under high levels of label noise.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Segmentation
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

Label noise is a critical problem in medical image segmentation, often arising from the inherent difficulty of manual annotation. Models trained on noisy data are prone to overfitting, which degrades their generalization performance. While a number of methods and strategies have been proposed to mitigate noisy labels in the segmentation domain, this area remains largely under-explored. The abstention mechanism has proven effective in classification tasks by enhancing the capabilities of Cross Entropy, yet its potential in segmentation remains unverified. In this paper, we address this gap by introducing a universal and modular abstention framework capable of enhancing the noise-robustness of a diverse range of loss functions. Our framework improves upon prior work with two key components: an informed regularization term to guide abstention behaviour, and a more flexible power-law-based auto-tuning algorithm for the abstention penalty. We demonstrate the framework's versatility by systematically integrating it with three distinct loss functions to create three novel, noise-robust variants: GAC, SAC, and ADS. Experiments on the CaDIS and DSAD medical datasets show our methods consistently and significantly outperform their non-abstaining baselines, especially under high noise levels. This work establishes that enabling models to selectively ignore corrupted samples is a powerful and generalizable strategy for building more reliable segmentation models. Our code is publicly available at https://github.com/wemous/abstention-for-segmentation.

Problem

• label noise
• medical image segmentation
• noise-robust learning
• abstention
• generalization

Innovation

• abstention
• noise-robust learning
• medical image segmentation
• adaptive regularization
• modular framework