M-GCN: A Multimodal Graph Convolutional Network to Integrate Functional and Structural Connectomics Data to Predict Multidimensional Phenotypic Characterizations
Keywords: Graph Convolutional Networks, Functional Connectomics, Structural Connectomics, Multimodal Integration, Phenotypic Prediction, Autism Spectrum Disorder
TL;DR: We propose a multimodal graph convolutional network to integrate functional and structural connectivity to predict multidimensional phenotypic characterizations.
Abstract: We propose a multimodal graph convolutional network (M-GCN) that integrates resting-state fMRI connectivity and diffusion tensor imaging tractography to predict phenotypic measures. Our specialized M-GCN filters act topologically on the functional connectivity matrices, as guided by the subject-wise structural connectomes. The inclusion of structural information also acts as a regularizer and helps extract rich data embeddings that are predictive of clinical outcomes. We validate our framework on 275 healthy individuals from the Human Connectome Project and 57 individuals diagnosed with Autism Spectrum Disorder from an in-house data to predict cognitive measures and behavioral deficits respectively. We demonstrate that the M-GCN outperforms several state-of-the-art baselines in a five-fold cross validated setting and extracts predictive biomarkers from both healthy and autistic populations. Our framework thus provides the representational flexibility to exploit the complementary nature of structure and function and map this information to phenotypic measures in the presence of limited training data.
Registration: I acknowledge that publication of this at MIDL and in the proceedings requires at least one of the authors to register and present the work during the conference.
Source Code Url: https://github.com/Niharika-SD/M-GCN
Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.
Data Set Url: https://www.humanconnectome.org/
Paper Type: both
Source Latex: zip
Primary Subject Area: Unsupervised Learning and Representation Learning
Secondary Subject Area: Integration of Imaging and Clinical Data