Brain Cortical Functional Gradients Predict Cortical Folding Patterns via Attention Mesh Convolution

Li Yang, Zhibin He, Tianyang Zhong, Changhe Li, Dajiang Zhu, Junwei Han, Tianming Liu, Tuo Zhang

Abstract: Research has shown a strong link between brain function and cortical folding using various imaging techniques and genetics. Understanding the functional roles of gyri and sulci in cortical folding patterns is crucial for insights into biological and artificial neural networks. However, the complex relationship, individual variations, and intricate brain function distribution pose challenges in developing a comprehensive theory and computational model. To address this, a new model leveraging brain functional gradients from fMRI data was developed to predict individual cortical folding maps. The model incorporates attention mesh convolution to account for spatial organization, showing superior performance compared to existing models. Discoveries indicate that less dominant functional gradients play a significant role in folding prediction, with cortical landmarks found on borders of activated regions. The results highlight the potential of tailored neural networks in enhancing the understanding of brain anatomy-function relationships.