Analysis of cortical and sub-cortical regions in autistic mr images using level set method and structure tensors

Autism is a complex disease that causes micro structural changes in brain due to delayed growth. In this work, the cortical regions of brain are extracted from T1 -weighted magnetic resonance mid-sagittal view slices. The images are skull stripped using edge based Level Set (LS) which is regularized by a signed distance function. This skull stripping technique is validated with the Brain extraction tool (BET). The extracted cortical regions are then analyzed by calculating the structure tensor matrix from the gradient of the skull stripped images. The anisotropy index derived from the tensor matrix is correlated with the clinical features such as verbal, full scale and performance Intelligent Quotients (IQ). The results show that the level set method is able to extract the cortical brain boundaries correctly with distinct edges. The structure tensor features extracted from the skull stripped images discriminates the control and autistic images. Also, the anisotropy index is negatively correlated with the performance IQ values of autistic subjects.