Analysis and evaluation of autistic brain MR images using Learning Vector Quantization and Support Vector Machines

Autism is a neuro-developmental disorder that retards the normal cognitive development of an affected person. It is prevalent in children below the age of five and is generally identified through the symptoms exhibited by them while they interact with the environment. This work focuses on the extraction of texture features for autistic and control subjects and validation is done using the neural classifiers, Learning Vector Quantization (LVQ) and Support Vector Machines (SVM). Six texture features namely, energy, entropy, contrast, inverse differential moment, directional moment, correlation and homogeneity were extracted for 15 autistic and 15 control groups through Gray Level Co-occurrence Matrix (GLCM). The system has been trained by subjecting these texture features using the LVQ and SVM. In order to ensure correctness of this mechanism, the validation has been done by employing the same techniques, where in LVQ gave a classification accuracy of 87.7% and SVM accounted 97.8% of classification accuracy.