Mutual Evidential Deep Learning for Medical Image Segmentation

2025-05-18

AI Summary

Existing semi-supervised medical image segmentation co-training methods suffer from model bias induced by low-quality pseudo-labels, as their averaging-based ensemble strategies ignore inter-pseudo-label reliability discrepancies. To address this, we propose a multi-architecture collaborative, evidence-driven framework: (1) heterogeneous networks generate complementary predictive evidence; (2) a class-aware evidence fusion mechanism enables reliability-weighted pseudo-label ensemble; and (3) asymptotic Fisher information modeling quantifies uncertainty to guide uncertainty-aware curriculum learning. This work is the first to deeply integrate Dempster–Shafer evidence theory with progressive curriculum learning, establishing a mutual-verification paradigm for pseudo-label generation. Evaluated on five mainstream medical imaging benchmarks, our method achieves state-of-the-art performance, significantly improving pseudo-label quality, model robustness, and segmentation accuracy on challenging cases.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Segmentation
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Existing semi-supervised medical segmentation co-learning frameworks have realized that model performance can be diminished by the biases in model recognition caused by low-quality pseudo-labels. Due to the averaging nature of their pseudo-label integration strategy, they fail to explore the reliability of pseudo-labels from different sources. In this paper, we propose a mutual evidential deep learning (MEDL) framework that offers a potentially viable solution for pseudo-label generation in semi-supervised learning from two perspectives. First, we introduce networks with different architectures to generate complementary evidence for unlabeled samples and adopt an improved class-aware evidential fusion to guide the confident synthesis of evidential predictions sourced from diverse architectural networks. Second, utilizing the uncertainty in the fused evidence, we design an asymptotic Fisher information-based evidential learning strategy. This strategy enables the model to initially focus on unlabeled samples with more reliable pseudo-labels, gradually shifting attention to samples with lower-quality pseudo-labels while avoiding over-penalization of mislabeled classes in high data uncertainty samples. Additionally, for labeled data, we continue to adopt an uncertainty-driven asymptotic learning strategy, gradually guiding the model to focus on challenging voxels. Extensive experiments on five mainstream datasets have demonstrated that MEDL achieves state-of-the-art performance.

Problem

• Addresses biases from low-quality pseudo-labels in semi-supervised medical segmentation
• Proposes evidential fusion and learning to improve pseudo-label reliability
• Enhances model focus on uncertain samples without over-penalizing mislabels

Innovation

• Different architecture networks for complementary evidence
• Class-aware evidential fusion for confident prediction synthesis
• Asymptotic Fisher information-based evidential learning strategy