Computer Vision News - March 2021

5 Autism classified by MRI people and, in some cases, they may find it hard to look after themselves. Creativity is commonly connectedwith ASD and this is one of the reasons (together with accepting our differences) that many organizations are working towards accepting this diversity in the professional and educational environment. Using imaging techniques, such as MRI, and in the field of neurobiology using biomarkers to compliment the clinical diagnosis, could lower the diagnostic costs, decrease the waiting time for the diagnosis and ideally achieve higher diagnostic accuracy. As noted in the paper, “it could also help to define endophenotypes that can guide genetic research and evaluation of pharmacotherapy”. Many neuroanatomical biomarkers have been proposed such as changes in hippocampal volume and shape, larger caudate nuclei, and the like, although some of the results may seem conflicting. The diversity of underlying causes and the phenotypic expression of the disorder is the reason for identification of biologically relevant endophenotypes for deconstructing ASD. Only in the last 10 years, automated segmentation software has allowed researchers to compare brains in large scale. Especially interesting is the fact that whole brain segmentation can be performed rather than a single area volumetry. Machine learning is also of special use here. Using data from magnetic resonance imaging, diffusion tensor imaging, and functional magnetic resonance imaging and models such as support vector machines (SVMs) allow for classification of those paradigms. The grey matter structure and white matter tract thickness and connectivity pattern, as well as microscopic cytoarchitecture, functional neurophysiological aspects do affect the expressed behavioral phenotype. All those parameters and ideally from different modalities are important to include in such approaches. This paper explores another route. Machine learning models can sometimes be more complicated for clinicians to deal with. Therefore, a multivariate statistical method for classification using brain magnetic resonance imaging segmentations is used. The classification of a clinical cohort and the method and implementation for the classification of autism is presented. Multivariate method The dataset is made from 45 adult males with 21 autistic cases and 21 typically