BrainNetDiff: Generative AI Empowers Brain Network Construction Via Multimodal Diffusion

Brain network analysis has emerged as a pivotal technology for gaining a deeper understanding of brain functions and disease mechanisms. However, existing network construction methods still have shortcomings in the learning of correlations between structural and functional brain imaging. In view of this, we introduce a novel method called BrainNetDiff, which combines a multi-head Transformer encoder to extract relevant features from fMRI and integrates a conditional latent diffusion model for brain network construction. By utilizing the conditional prompt and the attention mechanism, this method improves the accuracy and stability of brain network construction. To our knowledge, this represents the first work that employs diffusion for the fusion of multimodal brain imaging and brain network construction. We validate this framework by constructing brain networks on the public dataset ADNI. Experimental results demonstrate the effectiveness of the proposed method across the downstream disease classification tasks. This research provides a valuable reference for the processing of multimodal brain imaging and introduces a novel solution for brain network construction.