MultiCo3D: Multi-Label Voxel Contrast for One-Shot Incremental Segmentation of 3D Neuroimages

2025-03-09

AI Summary

To address semantic conflict among multiple labels and catastrophic forgetting in single-sample class-incremental segmentation of white matter tracts from 3D neuroimaging, this paper proposes the first contrastive learning framework tailored for multi-label voxels. Methodologically, it integrates uncertainty-aware knowledge distillation with a dynamic weighted multi-loss mechanism to achieve feature decoupling and joint optimization of old and new classes within the 3D U-Net encoder. Key contributions include: (1) extending voxel-wise contrastive learning to multi-label settings, mitigating semantic confusion caused by feature overlap between base and novel classes; (2) preserving discriminative knowledge of old classes via uncertainty-guided distillation; and (3) adaptively balancing segmentation accuracy and stability through loss weight adjustment. Evaluated on five incremental settings across the HCP and Pretto datasets, the method achieves an average segmentation accuracy improvement of 4.2–7.8 percentage points over state-of-the-art approaches.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Segmentation
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition

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
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

3D neuroimages provide a comprehensive view of brain structure and function, aiding in precise localization and functional connectivity analysis. Segmentation of white matter (WM) tracts using 3D neuroimages is vital for understanding the brain's structural connectivity in both healthy and diseased states. One-shot Class Incremental Semantic Segmentation (OCIS) refers to effectively segmenting new (novel) classes using only a single sample while retaining knowledge of old (base) classes without forgetting. Voxel-contrastive OCIS methods adjust the feature space to alleviate the feature overlap problem between the base and novel classes. However, since WM tract segmentation is a multi-label segmentation task, existing single-label voxel contrastive-based methods may cause inherent contradictions. To address this, we propose a new multi-label voxel contrast framework called MultiCo3D for one-shot class incremental tract segmentation. Our method utilizes uncertainty distillation to preserve base tract segmentation knowledge while adjusting the feature space with multi-label voxel contrast to alleviate feature overlap when learning novel tracts and dynamically weighting multi losses to balance overall loss. We compare our method against several state-of-the-art (SOTA) approaches. The experimental results show that our method significantly enhances one-shot class incremental tract segmentation accuracy across five different experimental setups on HCP and Preto datasets.

Problem

• Addresses multi-label segmentation in 3D neuroimages
• Enhances one-shot class incremental tract segmentation
• Reduces feature overlap between base and novel classes

Innovation

• Multi-label voxel contrast framework
• Uncertainty distillation for knowledge preservation
• Dynamic weighting of multi losses