Novel computational methods for higher order diffusion MRI in autism

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): The diagnosis of autism spectrum disorder (ASD) is currently based on behavior and developmental history of the child. With the development of advanced forms of diffusion-weighted magnetic resonance imaging (DW-MRI), it is expected that imaging will elucidate pathology-induced and neuro-developmental changes in white matter (WM) architecture, and provide diagnostic and predictive anatomical biomarkers. We aim at developing computational methods for processing and analysis of high angular resolution diffusion imaging data that has been fitted with higher order diffusion models (HOMs). Compared to the tensor model in diffusion tensor imaging (DTI), HOMs provide a much richer understanding of pathology-based connectivity changes in complex WM regions, as well as a quantification of the degree of abnormality of WM. These imaging measures when correlated with clinical measures of symptom severity will provide additional insight into the pathology and its progression, thus making this project very clinically significant. Understanding such complex WM regions is expected to aid in the study of ASD, deficits in which can be linked with WM abnormalities and disruptions in structural connectivity via fiber tracts. The advances in acquisition of data that can be fitted with HOMs in turn calls for novel automated tools for analyzing such data, as existing methods developed for tensors are inapplicable to HOMs. We propose to achieve this by the following specific aims: In Aim 1, we will define local and global measures from HOMs and use these to obtain a feature-based algorithm for deformable registration of HOM images preparing them for subsequent analysis. In Aim 2, we will develop and validate an integrated framework for population statistics of HOMs using a combination of voxel-based, manifold-based and tract-based analysis. In Aim 3, we will design high- dimensional multivariate pattern classifiers using HOM features, to obtain spatial patterns of brain abnormality and assign an abnormality to each brain. In Aim 4, we will apply the methods developed in Aims 1 - 3 to a large database of ASD patients and demographically balanced typically developing volunteers and identify patient-control differences and correlate with clinical ratings of symptom severity in patients. The quantification of patterns of group differences and connectivity disruptions are expected to provide insight into the deficits observed in autism such as impaired social interactions, impaired language and communication and stereotypical, restricted and repetitive behaviors. The use of HOMs that has never been attempted before in literature to study ASD, with most of the work limited to the analysis of anisotropy and diffusivity measures computed from DTI data. We expect that upon successful completion of the project, we have developed a general and comprehensive, mathematically consistent and computationally efficient processing and analysis paradigm for large population studies using HOMs that will help identify and quantify complex patterns of connectivity changes induced by pathology.