Understanding how large-scale functional brain networks reorganize during cognitive decline remains a central challenge in neuroimaging. While recent self-supervised models have shown promise for learning representations from resting-state fMRI, their internal mechanisms are difficult to interpret, limiting mechanistic insight. We propose BrainInterNet, a network-aware self-supervised framework based on masked reconstruction with cross-attention that explicitly models inter-network dependencies in rs-fMRI. By selectively masking predefined functional networks and reconstructing them from remaining context, our approach enables direct quantification of network predictability and interpretable analysis of cross-network interactions. We train BrainInterNet on multi-cohort fMRI data (from the ABCD, HCP Development, HCP Young Adults, and HCP Aging datasets) and evaluate on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, in total comprising 5,582 recordings. Our method reveals systematic alterations in the brain's network interactions under AD, including in the default mode, limbic, and attention networks. In parallel, the learned representations support accurate Alzheimer's-spectrum classification and yield a compact summary marker that tracks disease severity longitudinally. Together, these results demonstrate that network-guided masked modeling with cross-attention provides an interpretable and effective framework for characterizing functional reorganization in neurodegeneration.
Figures & tables
Figure 1: Overview of the proposed framework for modeling inter-network interactions in resting-state fMRI. Resting-state BOLD signals are parcellated using the DiFuMo atlas and divided into non-overlapping temporal-spatial patches. Patches are grouped by Yeo functional networks [ 24 ] , and those belonging to a selected target network are masked. The encoder processes unmasked tokens to produce contextualized embeddings, while a cross-attention decoder reconstructs the masked network using the context and position embeddings. Learned embeddings reveal separation by disease status, offering a simple metric for mapping neurodegeneration trajectories (bottom left). Meanwhile, the cross-attention decoder provides interpretability into how predictions are made, offering a proxy for inter-network brain dependencies (bottom right).
Figure 2: Longitudinal charting of Alzheimer’s disease progression using encoder embeddings. Left: Representative longitudinal trajectories of the encoder embedding norm ( ℓ2 ) for individual subjects, including stable CN, stable MCI, stable AD, and converters who transition from MCI to AD across sessions. Faint background points show embedding norms for all subjects and sessions. Right: Group-level comparison of embedding norms for stable CN, MCI, and AD subjects, showing progressive separation with disease severity (mean ± s.e.m.). *** indicates p<0.001 (two-sided Welch’s t-test). Together, these analyses illustrate that the learned encoder representations capture both group-level differences and subject-specific trajectories associated with cognitive decline.
Figure 3: Inter-network predictability and disease-related changes in functional coupling . Control: Baseline inter-network predictability in cognitively normal (CN) subjects. Chord diagrams depict cross-network attention-based contributions during reconstruction, where edge thickness indicates the relative influence of source networks (left) on target network prediction (right). Bar plots report network-wise predictability measured by Pearson correlation between reconstructed and true signals. MCI - Control: Changes in inter-network contributions from CN to MCI. Edges and bars represent differences in attention-derived source dependencies relative to CN. AD - Control: Analogous changes in inter-network contributions from CN to AD. Red indicates increased reliance on a given source network for reconstruction, while blue indicates decreased reliance. Together, these panels illustrate progressive reorganization of functional dependencies from cognitive impairment and Alzheimer’s disease.
Method
Mask
CrossAttn
BAcc (%)
F1 (%)
AUC
Baselines
BrainLM [ 7 ]
Rand
✗
68.78
68.70
0.776
Brain-JEPA [ 13 ]
Rand
✗
51.85
51.10
0.620
BrainGNN [ 19 ]
NA
✗
40.89
40.22
0.553
Random
NA
✗
33.65
31.50
0.503
Ablations
MAE [ 18 ]
Rand
✗
69.31
71.86
0.849
Ours - Net
Rand
✓
48.26
43.23
0.520
Table 1: Performance comparison on 3-way Alzheimer’s disease spectrum classification on ADNI. Reported metrics including balanced accuracy (BAcc), F1-score, and AUC-ROC. We test four baselines, and three ablations: a vanilla MAE, replacing cross-attention with self-attention, as well as replacing the network-based masking (denoted Net in the table) with random masking (denoted Rand). All models are finetuned.
Accurate identification of Alzheimers disease (AD) using resting-state functional magnetic resonance imaging (rs-fMRI) remains challenging due to the high dimensionality, noise, and complex inter-regional dependencies inherent in functional brain connectivity, which limit the effectiveness of traditional approaches based on handcrafted connectivity features or conventional machine learning models. In this work, we present an attention-based deep learning framework for Alzheimers disease classification that operates directly on rs-fMRI functional connectivity matrices by treating brain regions as tokens and employing a Transformer-inspired self-attention mechanism to model long-range and global functional dependencies across distributed brain networks. The proposed framework learns discriminative functional representations without reliance on manual feature engineering and is evaluated on a longitudinal cohort from the Alzheimers Disease Neuroimaging Initiative (ADNI) comprising cognitively normal and Alzheimers disease subjects with multiple visits. A subject-wise evaluation protocol is adopted to prevent information leakage across visits, and class-weighted optimization is incorporated to address mild class imbalance. Experimental results for binary AD versus cognitively normal classification demonstrate that the proposed attention- based rs-fMRI model achieves an accuracy of 88.95% and a ROC-AUC of 0.90, along with a favorable precision-recall balance, highlighting the effectiveness of self-attention-driven functional connectivity modeling as a robust and interpretable approach for Alzheimers disease detection using resting-state fMRI.
Harshiddhi Pathak, Gowtham Reddy N, Mrinal Acharya +1
School of Medical Science and Technology · Department of Computer Science and Engineering at the Indian Institute of Technology Gandhinagar · Dr B C Roy Multi Speciality Medical Research Centre, Indian Institute of Technology Kharagpur, India
Masked autoencoders (MAEs) have recently shown promise for self-supervised representation learning of resting-state brain functional connectivity (FC). However, a fundamental question remains unresolved: how should FC matrices be tokenized to align with the intrinsic modular organization of large-scale brain networks? Existing approaches typically adopt region-centric or graph-based schemes that treat FC as structurally homogeneous elements and overlook the large-scale network brain organization. We introduce NERVE (Network-Aware Representations of Brain Functional Connectivity via Bilinear Tokenization), a self-supervised learning framework that redefines FC tokenization by partitioning FC matrices into patches of intra- and inter-network connectivity blocks. Unlike image-based MAE, where fixed-size patches share a common tokenizer, FC patches defined by network pairs are heterogeneous in size and correspond to distinct functional roles. To resolve this problem, NERVE embeds FC patches through a novel structured bilinear factorization. This formulation preserves network identity and reduces parameter complexity from quadratic to linear scaling in the number of networks. We evaluate NERVE across three large-scale developmental cohorts (ABCD, PNC, and CCNP) for behavior and psychopathology prediction. Compared to structurally agnostic MAE variants and graph-based self-supervised baselines, the proposed network-aware formulation yields more stable and transferable representations, particularly in cross-cohort evaluation. Ablation studies confirm that the proposed bilinear network embedding and anatomically grounded parcellation are critical for performance. These findings highlight the importance of incorporating domain-specific structural priors into self-supervised learning for functional connectomics. Code is available at: https://github.com/leomlck/NERVE.
Leo Milecki, Qingyu Hu, Bahram Jafrasteh +2
Department of Radiology, Weill Cornell Medicine, New York, NY, USA. · School of Electrical and Computer Engineering, Cornell University and Cornell Tech, New York, NY, USA.
Self-supervised pretraining is promising for large-scale neuroimaging, yet the impact of region-aware masking and hybrid sequence modeling remains underexplored. In this work, we introduce Rhamba, a region-aware pretraining framework that integrates anatomically guided masking with hybrid Attention-Mamba architectures for resting state functional magnetic resonance imaging (fMRI) analysis. Models were pretrained on the ABIDE dataset using region-aligned patch embeddings and three masking strategies (Any, Majority, and Pure) with increasing spatial specificity. We evaluated four architectural variants: a Mamba only model, an Alternate architecture with interleaved Mamba and Attention blocks, and two hybrid encoder-decoder configurations (Attention-Mamba (AM) and Mamba-Attention (MA)). The pretrained models were fine-tuned on downstream classification tasks using the COBRE and ADHD-200 datasets for schizophrenia and attention-deficit/hyperactivity disorder discrimination. We employed Integrated Gradients, an explainable AI method, to identify the brain regions contributing to model predictions. Masking strategy strongly influenced reconstruction behavior, with reconstruction loss following a consistent ordering (Any > Majority > Pure). However, this trend did not directly translate into downstream performance, where differences were modest and dataset-dependent. The hybrid architecture with the MA configuration achieved the highest average AUROC across both datasets, and Rhamba outperformed state-of-the-art methods in comparative evaluation. Region-wise analysis showed that peak performance depends on the interaction between masking strategy and architecture rather than a single dominant configuration. Overall, Rhamba offers a flexible framework for balancing interpretability, scalability, and performance in large-scale fMRI representation learning.