MetaSSL: A General Heterogeneous Loss for Semi-Supervised Medical Image Segmentation

2025-09-01

AI Summary

In medical image segmentation, semi-supervised methods often overlook label noise and the heterogeneous value of unlabeled pixels. To address this, we propose MetaSSL—a generic framework that jointly models prediction consistency and uncertainty within the loss function, enabling spatially heterogeneous weighting. Specifically, MetaSSL dynamically partitions predictions into four regions via adaptive thresholds to distinguish reliable from ambiguous pixel-level predictions, thereby enabling fine-grained, pixel-wise supervision. Importantly, MetaSSL is architecture-agnostic and seamlessly integrates with mainstream semi-supervised frameworks (e.g., Mean Teacher, FixMatch) as a plug-and-play module. Extensive experiments across multiple medical segmentation benchmarks demonstrate that MetaSSL significantly improves performance while exhibiting strong robustness to label noise, efficient utilization of unlabeled data, and superior generalizability.

Technology Category

• Machine Learning: Semi-Supervised Learning
• Computer Vision: Segmentation
• Search and Optimization: Metareasoning and Metaheuristics

Application Category

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

Abstract

Semi-Supervised Learning (SSL) is important for reducing the annotation cost for medical image segmentation models. State-of-the-art SSL methods such as Mean Teacher, FixMatch and Cross Pseudo Supervision (CPS) are mainly based on consistency regularization or pseudo-label supervision between a reference prediction and a supervised prediction. Despite the effectiveness, they have overlooked the potential noise in the labeled data, and mainly focus on strategies to generate the reference prediction, while ignoring the heterogeneous values of different unlabeled pixels. We argue that effectively mining the rich information contained by the two predictions in the loss function, instead of the specific strategy to obtain a reference prediction, is more essential for SSL, and propose a universal framework MetaSSL based on a spatially heterogeneous loss that assigns different weights to pixels by simultaneously leveraging the uncertainty and consistency information between the reference and supervised predictions. Specifically, we split the predictions on unlabeled data into four regions with decreasing weights in the loss: Unanimous and Confident (UC), Unanimous and Suspicious (US), Discrepant and Confident (DC), and Discrepant and Suspicious (DS), where an adaptive threshold is proposed to distinguish confident predictions from suspicious ones. The heterogeneous loss is also applied to labeled images for robust learning considering the potential annotation noise. Our method is plug-and-play and general to most existing SSL methods. The experimental results showed that it improved the segmentation performance significantly when integrated with existing SSL frameworks on different datasets. Code is available at https://github.com/HiLab-git/MetaSSL.

Problem

• Reducing annotation costs for medical image segmentation models
• Addressing noise in labeled data and unlabeled pixel heterogeneity
• Improving semi-supervised learning through spatially adaptive loss weighting

Innovation

• Heterogeneous loss weighting by uncertainty
• Adaptive threshold for confident predictions
• Plug-and-play framework for SSL methods