Madhura Ingalhalikar1, Drew Parker1,
Timothy P. L. Roberts2, Ragini Verma1
1Section of Biomedical
Image Analysis, Department of Radiology, University of Pennsylvania,
Philadelphia, PA, United States; 2Lurie Family Foundations MEG
Imaging Center, Childerns Hospital of Philadelphia, Philadelphia, PA
This work presents a paradigm for generating a pathology based marker of language impairment (LI) in Autism Spectrum Disorder (ASD). This is achieved by combining MEG and DTI features in a pattern classifier. These classifiers, in addition to group separation have a potential to quantify the degree of LI by assigning an abnormality score to each subject. Furthermore, the ranking of features gives a physiological insight into the pathology. A 3-way classification between LI-ASD, non-LI-ASD and typically developing (TD) controls was achieved with an average accuracy of 71.69% providing better understanding of LI than just using individual modalities.