Brain disorders involve abnormalities spanning multiple biological and phenotypic scales that require the integration of multiple data sources, including neuroimaging and structured clinical data, to achieve accurate diagnosis. However, effectively integrating heterogeneous imaging modalities with tabular clinical data remains challenging due to fundamental differences in data structure, dimensionality, and semantics. Existing fusion strategies often restrict fusion to a single network layer, hindering the progressive integration of complementary features across multiple layers and limiting the sufficiency and flexibility of interactions between modalities. To address these limitations, we propose an Adaptive Tabular-guided Layerwise Alignment and Synthesis Network (ATLAS-Net) for effective multimodal neuroimaging and tabular data integration. Specifically, ATLAS-Net introduces an adaptive layerwise alignment (ALA) module that progressively aligns and fuses features from imaging branches through attention-driven interactions at multiple network layers, enabling continuous and sufficient cross-modality information exchange. In addition, we design a tabular-adaptive parameterization (TAP) module, which incorporates structured clinical variables as conditional priors to dynamically generate network parameters and modulate intermediate feature channels, thereby allowing tabular information to directly guide image feature extraction throughout the network hierarchy. Extensive experiments on 8 brain disorders demonstrate that ATLAS-Net consistently outperforms state-of-the-arts in classification performance, cross-dataset generalization, and robustness to realistic spatial noise perturbations. The code is available at https://github.com/liangchuang11/ATLAS-Net.git.
Chuang Liang, T. Adalı, Jing Sui et al.· Proceedings of the 32nd ACM...· 0 citations
The emergence of foundation models has revolutionized neuroimaging analysis, offering universal representations for decoding complex brain functions. However, existing fMRI-based foundation models predominantly rely on standard group-level atlases neglecting inter-subject heterogeneity. They often operate on voxel-wise time series with inherent low signal-to-noise ratios (SNR) and employ single-objective pre-training strategies that fail to capture fine-grained local structures and global discriminative representations. To bridge these gaps, we present BrainNetGFM, a novel graph-based foundation model for brain network construction that integrates individualized geometry with joint self-supervised learning, pre-trained on a massive dataset of 44,476 samples. First, we introduce an individualized brain atlas construction method utilizing region growing on functional priors to ensure precise alignment of subject-specific functional topology. Brain network graphs integrating functional connectivity strength with individualized spatial geometry is constructed. Second, we propose a joint self-supervised learning (SSL) framework that synergizes generative and contrastive paradigms. Specifically, we combine graph masked autoencoders (GMAE) to reconstruct masked node and edge features for learning fine-grained local structures, alongside multi-level graph contrastive learning (GCL) spanning both node and graph levels to enforce global discriminative representations. Extensive experiments on downstream tasks including the diagnosis of 8 brain disorders, age and cognition prediction, demonstrate that BrainNetGFM outperforms 9 state-of-the-art methods. Moreover, BrainNetGFM exhibits superior interpretability, identifying neurologically meaningful network biomarkers for clinical analysis. The code is publicly available at https://github.com/BrainLabA/BrainNetGFM.
Chunzhi Zhao, T. Adalı, Jing Sui et al.· Proceedings of the 32nd ACM...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
