Jeong-Won Jeong PhD, Presenter: Nothing to Disclose

Harry T. Chugani MD, Abstract Co-Author: Nothing to Disclose

Diane C Chugani PhD, Abstract Co-Author: Research Consultant, Shire plc

Vijay Narayan Tiwari MD, PhD, Abstract Co-Author: Nothing to Disclose

To develop imaging biomarkers that differentiate autistic children with normal cognitive function (high functioning autism, HFA) from those with impaired cognitive function (low functioning autism, LFA), as well as from typically developing children (TD), we examined autism-related white matter diffusivity changes using a tract-based spatial statistics (TBSS) analysis. The robustness of obtained parameters was tested via a multi-parametric classification approach based on a support vector machine (SVM).

14 children with HFA (mean age: 5.7±2 years; 10 males), 14 children with LFA (mean age: 5.5±1.7 years; 10 males), and 14 typically developing children (TD, mean age: 7.1±3.5 years; 9 males) underwent diffusion tensor imaging to create fractional anisotropy (FA) and spherical diffusivity (SD) images. In a first step, TBSS analysis was applied for three different group comparisons of each diffusivity image, "TD vs. autism (LFA+HFA)" and "TD vs. LFA". We identified the regions of interest (ROI) showing significant changes in HFA+LFA and LFA groups, compared to the TD group. These ROIs were used to construct feature vectors for two SVM classifiers, SVM #1 to differentiate the HFA+LFA group from the TD, and SVM #2 to differentiate the LFA and HFA groups. For each SVM classification, the accuracy of correct classification was evaluated using "hold-out" cross validation.

Compared with the TD group, both HFA and LFA groups showed significantly reduced FA in bilateral posterior corona radiata, right anterior corona radiata, and corpus callosum. Compared with the HFA, the LFA showed significantly increased SD in right fornix and right cerebellum. Using the above ROIs, the proposed SVM classification provided high accuracy, 933 % for "SVM #1: TD vs. autism" and 847 % for "SVM #2: HFA vs. LFA".

Our study revealed that FA in bilateral posterior corona radiata, right anterior corona radiata, and corpus callosum can differentiate children with autism from TD children, and children with LFA can be distinguished from those with HFA by comparing the value of SD in right fornix and right cerebellum.

The findings in the present study could be used as imaging biomarkers for differentiating HFA and LFA and TD.

Jeong, J, Chugani, H, Chugani, D, Tiwari, V, Diffusion Tensor MRI Biomarker for Differentiating Children with High and Low Functioning Autism.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11012201.html