CRAF-Net: A Fine-Grained Cross-Channel Attention Network for Preoperative Microvascular Invasion Grading in
Zebang Zhong1, Xiao Luo1, Daoying Geng1,2,3
1College of Biomedical Engineering, Qidong-Fudan Innovative Institute of Medical Sciences, Fudan University, Shanghai, China.
None:
To address the clinical challenge of preoperative microvascular invasion (MVI) grading in hepatocellular carcinoma (HCC), this paper proposes a cross-channel attention fine-grained Network (CRAF-Net) using dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). Current methods often suffer from fragmented representations and suboptimal channel weighting, obscuring key features and reducing accuracy. CRAF-Net addresses these challenges through three innovations: (1) a Dense-NeXt encoder that enhances 3D hierarchical feature reuse; (2) a cross-channel attention (CCA) Module that adaptively weights multi-phase MRI sequences to emphasize tumor microenvironment (TME) biomarkers; and (3) a dual-branch disentangled network with a multi-classifier fusion Module that separates morphological MVI risk assessment from severity grading, alleviating inter-class ambiguities. Evaluated on a primary cohort of 472 HCC cases, CRAF-Net achieves an 84.8% three-category accuracy and a macro-average AUC of 0.950. Crucially, on an independent external cohort (n = 118), it demonstrates robust cross-center generalizability with 81.4% accuracy and a 0.923 AUC. Decision Curve analyses validate its optimal net benefit for clinical decision-making. Operating efficiently (0.6 s/case), the model satisfies real-time clinical requirements. Finally, survival analysis demonstrates strong concordance between predicted grades and pathological outcomes (5-year overall survival: M0 = 81.4%, M1 = 69.5%, M2 = 43.5%). CRAF-Net serves as a robust auxiliary framework for personalized HCC management.
