Cortical Surface Diffusion Generative Models

Zhenshan Xie, Simon Dahan, Logan Zane John Williams, M. Jorge Cardoso, Emma C. Robinson

Keywords: Cortical Surfaces, Diffusion Models, Generative Models

Abstract: Cortical surface analysis is gaining prominence as a sensitive tool for studying complex neuropsychiatric disorders. However, patterns of cortical organization are complex and highly variable across individuals, challenging classical approaches for analysis that rely on diffeomorphic image registration. This leads to an urgent need for better methods to model brain development and diverse variability inherent across different individuals. Traditional vision diffusion models have shown effectiveness in generating high-resolution and realistic natural images, which makes them particularly suited to addressing cortical surface problems, where the features of interest are subtle and highly variable across individuals. In this work, we first proposed a novel diffusion model for the generation of cortical surface metrics, using modified surface vision transformers as the principal architecture. We validate our method in the developing Human Connectome Project (dHCP) with results suggesting that our model demonstrates excellent performance in capturing the intricate details of evolving cortical surfaces - generating high-quality realistic samples of cortical surfaces conditioned on postmenstrual age (PMA) at scan.

Submission Number: 30