DCMM-Transformer: Degree-Corrected Mixed-Membership Attention for Medical Imaging
This addresses the need for more effective and interpretable medical image analysis tools for clinicians and researchers, though it is incremental as it builds on prior work like SBM-Transformer.
The paper tackled the problem of standard Vision Transformers failing to exploit latent anatomical groupings in medical images by proposing DCMM-Transformer, which incorporates a Degree-Corrected Mixed-Membership model as an additive bias in self-attention, resulting in superior performance and enhanced interpretability across diverse medical imaging datasets.
Medical images exhibit latent anatomical groupings, such as organs, tissues, and pathological regions, that standard Vision Transformers (ViTs) fail to exploit. While recent work like SBM-Transformer attempts to incorporate such structures through stochastic binary masking, they suffer from non-differentiability, training instability, and the inability to model complex community structure. We present DCMM-Transformer, a novel ViT architecture for medical image analysis that incorporates a Degree-Corrected Mixed-Membership (DCMM) model as an additive bias in self-attention. Unlike prior approaches that rely on multiplicative masking and binary sampling, our method introduces community structure and degree heterogeneity in a fully differentiable and interpretable manner. Comprehensive experiments across diverse medical imaging datasets, including brain, chest, breast, and ocular modalities, demonstrate the superior performance and generalizability of the proposed approach. Furthermore, the learned group structure and structured attention modulation substantially enhance interpretability by yielding attention maps that are anatomically meaningful and semantically coherent.