Analysis of under-connectivity in Autism using the minimum spanning tree: application on large multi-site dataset

Autism spectrum disorder is a neurodevelopmental disorder that shows increasing occurrence. With no cure nor clear information about its causes, some works are turning toward the brain connectivity to extract patterns that will help in the early diagnosis. However, small data present a drawback that prevent final conclusions. In this light, this paper proposes a new approach to study the connections of an autistic brain, using the minimum spanning tree (MST). The MST permits to extract weak connections and thus, to verify the under-connectivity deficit reported in Autism. Once they are extract, the weakest connections are eliminated from the connectivity matrices of Rs-fMRI images in order to enhance their impact. The resultant matrices are then used for classification with SVM. The approach was tested on the large and heterogeneous ABIDE dataset. The results brought out a pattern that confirms the under-connectivity deficit in Autism.