gnn-visual-decoding-brain-network

name	gnn-visual-decoding-brain-network
description	Graph Neural Network approach for decoding visual category representations from large-scale brain functional networks using 7T fMRI data.
version	1.0.0
author	Research Synthesis
license	MIT
metadata	{"hermes":{"tags":["brain-decoding","gnn","fmri","visual-processing","brain-network","cognitive-neuroscience"],"source_paper":"Decoding Functional Networks for Visual Categories via GNNs (arXiv:2603.28931v1)"}}

GNN-Based Visual Category Decoding from Brain Networks

Overview

This approach uses Graph Neural Networks to decode visual category representations from high-resolution 7T fMRI data (Natural Scenes Dataset). By modeling brain functional connectivity as a graph and learning category-specific network patterns, the method links perception to cortical organization.

Core Concepts

Graph Construction

• Nodes: Brain regions/voxels from fMRI
• Edges: Functional connectivity (correlation/coherence between regions)
• Features: BOLD signal patterns during visual stimulus presentation

GNN Architecture

• Message passing over functional brain network
• Category-specific attention over brain regions
• Hierarchical aggregation from local to global network patterns

Implementation Pattern

class BrainNetworkGNN(nn.Module):
    def __init__(self, n_regions, hidden_dim, n_categories):
        super().__init__()
        self.node_encoder = nn.Linear(n_timepoints, hidden_dim)
        self.gcn1 = GraphConv(hidden_dim, hidden_dim)
        self.gcn2 = GraphConv(hidden_dim, hidden_dim)
        self.classifier = nn.Sequential(
            nn.Linear(hidden_dim, hidden_dim), nn.ReLU(),
            nn.Linear(hidden_dim, n_categories)
        )
    
    def forward(self, bold_signals, adj_matrix):
        h = self.node_encoder(bold_signals)
        h = self.gcn1(h, adj_matrix).relu()
        h = self.gcn2(h, adj_matrix).relu()
        h = h.mean(dim=0)  # Graph-level readout
        return self.classifier(h)

Applications

• Visual cortex mapping and analysis
• Brain-computer interfaces for visual perception
• Cognitive neuroscience research
• Neuroimaging-based category decoding

Activation Keywords

• brain network decoding, GNN fMRI analysis, visual category decoding, functional connectivity graph, 7T fMRI analysis, Natural Scenes Dataset, 脑网络解码, 图神经网络脑成像

References

• Decoding Functional Networks for Visual Categories via GNNs
• Authors: Shira Karmi, Galia Avidan, Tammy Riklin Raviv
• Published: 2026-03-30
• arXiv: https://arxiv.org/abs/2603.28931v1