Learnable Pooling in Graph Convolution Networks for Brain Surface Analysis

TL;DR

This paper introduces a learnable pooling method for graph convolutional networks that effectively processes complex brain surface data, improving analysis accuracy across multiple neuroscience prediction tasks.

Contribution

It presents a novel spectral embedding-based pooling strategy that adapts to multiple surface data, enhancing flexibility and performance in brain surface analysis.

Findings

Outperforms existing pooling methods in brain surface tasks

Improves state-of-the-art results in multiple prediction benchmarks

Demonstrates versatility across diverse neuroscience applications

Abstract

Brain surface analysis is essential to neuroscience, however, the complex geometry of the brain cortex hinders computational methods for this task. The difficulty arises from a discrepancy between 3D imaging data, which is represented in Euclidean space, and the non-Euclidean geometry of the highly-convoluted brain surface. Recent advances in machine learning have enabled the use of neural networks for non-Euclidean spaces. These facilitate the learning of surface data, yet pooling strategies often remain constrained to a single fixed-graph. This paper proposes a new learnable graph pooling method for processing multiple surface-valued data to output subject-based information. The proposed method innovates by learning an intrinsic aggregation of graph nodes based on graph spectral embedding. We illustrate the advantages of our approach with in-depth experiments on two large-scale benchmark datasets. The flexibility of the pooling strategy is evaluated on four different prediction tasks, namely, subject-sex classification, regression of cortical region sizes, classification of Alzheimer's disease stages, and brain age regression. Our experiments demonstrate the superiority of our learnable pooling approach compared to other pooling techniques for graph convolution networks, with results improving the state-of-the-art in brain surface analysis.