Interpretable Biomedical Large Models: From Multimodal Diagnosis to Molecular Mechanism Interpretation of Complex Diseases

Biomedical large models have shown strong performance in disease diagnosis using medical images, electrophysiological signals (e.g., EEG), and multi-omics data. However, their clinical translation remains limited by two persistent gaps: poor interpretability of model decisions, and a disconnection between model outputs and underlying molecular pathogenesis. A model that predicts a diagnosis with high accuracy but cannot explain why, or cannot be linked back to a biological mechanism, offers limited value to clinicians and researchers seeking actionable insight.This Research Topic focuses on interpretable biomedical large models spanning medical imaging, EEG-based brain signal analysis, and multi-omics data integration. We prioritize contributions that go beyond diagnostic performance metrics (accuracy, AUC, etc.) and instead interpret model outputs in terms of molecular, cellular, and pathological mechanisms — turning black-box predictions into mechanistic hypotheses that can inform biomarker discovery and therapeutic targeting.We welcome submissions across disease areas including but not limited to oncology, neurodegeneration, and neuropsychiatric disorders, and across article types: original research, methods papers, systematic reviews, perspectives, and FAIR² Data.• Foundation models for medical imaging• EEG foundation models and brain signal analysis• Computer-aided diagnosis of disease• Mechanisms, SHAP, counterfactual reasoning• Multimodal fusion across imaging, EEG, transcriptomics, proteomics• AI-driven discovery of molecular mechanisms and therapeutic targets Topic editors: Qiankun Zuo, Hongzhi Kuai, Jin Hong, Jia Guo Source: https://www.frontiersin.org/research-topics/85819/interpretable-biomedical-large-models-from-multimodal-diagnosis-to-molecular-mechanism-interpretation-of-complex-diseases