Generalising the HCP multimodal cortical parcellation to UK Biobank using geometric deep learning

Logan Zane John Williams, Simon Dahan, Tanya Poppe, Yourong Guo, Matthew Glasser, Fidel Alfaro-Almagro, Mohamed A. Suliman, Anderson Winkler, Timothy Coalson, David Van Essen, Stephen M. Smith, Emma C. Robinson

Keywords: geometric deep learning, image segmentation, UK Biobank, human connectome project, cortical surfaces, image registration

TL;DR: This paper uses geometric deep learning to generate individualised HCP-style multimodal parcellations in UK Biobank

Abstract: The Human Connectome Project Multimodal Parcellation (HCP\_MMP1.0) provides a robust \textit{in vivo} map of the cerebral cortex, which demonstrates variability in structure and function that cannot be captured through diffeomorphic image registration alone. The HCP successfully employed a fully-connected neural network architecture to capture this variation, however it is unclear whether this approach generalises to other datasets with less rich imaging protocols. In this paper we propose and validate a novel geometric deep learning framework for generating individualised HCP\_MMP1.0 parcellations in UK Biobank (UKB), an extremely rich resource that has led to numerous breakthroughs in neuroscience. To address substantial differences in image acquisition (for example, 6 minutes of resting-state fMRI per subject for UKB vs. 60 minutes per subject for HCP), we introduce a multi-step learning procedure including pretraining with a convolutional autoencoder. Compared to a fully-connected baseline, our proposed framework improved average detection rate across all areas by 10.4\% (99.1\% vs 88.7\%), and detection of the worst performing area by 51.0\% (80.9\% vs. 29.9\%). Importantly, this was not a result of the framework predicting one consistent parcellation across subjects, as visual inspection indicated that our method was sensitive to atypical cortical topographies. Code and trained models will be made available at \url{https://www.github.com}.

Primary Subject Area: Geometric Deep Learning

Secondary Subject Area: Application: Neuroimaging

Registration Requirement: Yes

Visa & Travel: Yes

Read CFP & Author Instructions: Yes

Originality Policy: Yes

Single-blind & Not Under Review Elsewhere: Yes

LLM Policy: Yes

Submission Number: 270