SurfGNN: A robust surface-based prediction model with interpretability for coactivation maps of spatial and cortical features

TL;DR

This work considers the cortical surface mesh as a sparse graph and proposes an interpretable prediction model-Surface Graph Neural Network (SurfGNN), which outperforms all existing state-of-the-art methods and generates feature-level activation maps, indicating its capability to identify robust regional variations in different morphometric contributions for prediction.

Abstract

Current brain surface-based prediction models often overlook the variability of regional attributes at the cortical feature level. While graph neural networks (GNNs) excel at capturing regional differences, they encounter challenges when dealing with complex, high-density graph structures. In this work, we consider the cortical surface mesh as a sparse graph and propose an interpretable prediction model-Surface Graph Neural Network (SurfGNN). SurfGNN employs topology-sampling learning (TSL) and region-specific learning (RSL) structures to manage individual cortical features at both lower and higher scales of the surface mesh, effectively tackling the challenges posed by the overly abundant mesh nodes and addressing the issue of heterogeneity in cortical regions. Building on this, a novel score-weighted fusion (SWF) method is implemented to merge nodal representations associated with each cortical feature for prediction. We apply our model to a neonatal brain age prediction task using a dataset of harmonized MR images from 481 subjects (503 scans). SurfGNN outperforms all existing state-of-the-art methods, demonstrating an improvement of at least 9.0% and achieving a mean absolute error (MAE) of 0.827+0.056 in postmenstrual weeks. Furthermore, it generates feature-level activation maps, indicating its capability to identify robust regional variations in different morphometric contributions for prediction.

Figures (6)

• Figure 1: Overview of the proposed network architecture and its key modules. N: the number of cortical features for each subject, input into SurfGNN. The model showcased above operates on an input mesh resolution of 5,124 nodes, necessitating two TSL structures for each cortical feature. For node number of 81,924, 20,484, 1,284, the model requires four, three and one TSL structures, respectively.
• Figure 2: Multi-resolution mesh maps on cerebral cortical surface. N: the number of nodes on the mesh.
• Figure 3: The scatter plot of the predicted brain ages and postmenstrual ages on the two cohorts with each input sparse graph consisting of 5,124 nodes. R: the correlation coefficient between the two axes for each cohort.
• Figure 4: Comparison of prediction performance of SurfGNN containing different numbers of TSL structures, corresponding to distinct resolutions of output sparse graphs after all the TSL structures and also inputs of RSL structure.
• Figure 5: Comparison between our model and two post-hoc approaches on the spatial activation maps for the three cortical features. The circles indicate areas of higher response within each feature. The various colors for each surface denote differences in qualitative importance. Maps from different approaches or distinct cortical features of the same approach are not directly comparable in terms of values.