29904
Machine Learning Analysis of White Matter Connectome Edge Density: A Path Towards Imaging Biomarkers for Autism Spectrum Disorders
Neuroimaging studies, using Diffusion tensor imaging (DTI) and fiber tractography have found white matter network underconnectivity in many children with the social communication challenges indicative of Autism Spectrum Disorders (ASD). Thus, the structural connectome, representing the whole-brain network of macro-scale white matter connectivity, has emerged during the past decade as a powerful formalism for the study of neurological and psychiatric diseases, including ASD. However, to date there are no studies of ASD examining regional connectomic properties within the white matter. Edge Density Imaging (EDI) has recently been introduced as a framework to represent the anatomic embedding of these white matter connectome edges. In EDI, the edges or links of the white matter connectome – from probabilistic tractography – are constrained to network nodes based on standard atlas parcellation of the cortical and deep gray matter nuclei. Machine learning analyses are also gaining popularity for pattern recognition and development of classification (or regression) models based on multidimensional data. These algorithms seem particularly suitable for devising classifiers based on multitude of variables extracted from diffusion and connectivity maps.
Objectives:
In this study, we compared the white matter connectome and microstructure between children with ASD and typically developing children (TDC) using voxel-wise analysis. Then, we applied different machine-learning algorithms for identification of ASD based on the white matter tract-based average Edge Density (ED) and conventional DTI metrics to determine the differences that were most associated with the clinical condition.
Methods:
We examined the structural connectome of children with (n=14) and without ASD (n=33) using tractography-based Edge Density Imaging (EDI); and then applied machine leaning algorithms to identify boys (8-12 years) with ASD based on EDI patterns. The Edge Density (ED) maps were computed from probabilistic streamline tractography applied to high angular resolution diffusion imaging (HARDI). Tract-Based Spatial Statistics (TBSS) was used for voxel-wise comparison and coregistration of ED maps in addition to conventional DTI metrics of Fractional Anisotropy (FA), Mean Diffusivity (MD), and Radial Diffusivity (RD). Tract-based average DTI/connectome metrics were calculated and used as input for different machine learning models: naïve Bayes, random forest, support vector machines (SVM), neural networks. For these models, cross-validation was performed with stratified random sampling (×1000) of the cohort into training and validation datasets. The average accuracy among validation samples was calculated.
Results:
In voxel-wise analysis, the body and splenium of corpus callosum, bilateral superior and posterior corona radiata, and left superior longitudinal fasciculus showed significantly lower ED in children with ASD. Overall, machine-learning models using tract-based EDI metrics had better performance in identification of children with ASD compared to those using FA, MD, and RD. The EDI-based random forest models had greater average accuracy (75.3%), specificity (97.0%), and positive predictive value (81.5%), whereas EDI-based polynomial SVM had greater sensitivity (51.4%), and negative predictive values (77.7%).
Conclusions:
In conclusion, we found reduced number of connectome edges in the posterior white matter tracts of children with ASD; and demonstrated the feasibility of connectome-based machine-learning algorithms in identification of children with ASD.