Autism Detection Using Surface and Volumetric Morphometric Feature of sMRI with Machine Learning Approach

Abstract

Brain imaging has played a very crucial role in the detection of various brain disorders. Among many brain imaging modalities, Magnetic Resonance Imaging (MRI) has proven its importance due to its detailed information regarding the insight of the brain. Autism Spectrum Disorder (ASD) has emerged as a very serious brain disorder due to its late detection among people. It comprises symptoms that are generally ignored, and this creates the urgency for its early detection. This work puts forward the method for the detection of ASD utilizing Machine Learning (ML) with the features extracted from sMRI (Structural Magnetic Resonance Imaging). Surface morphometric and volumetric morphometric features have been utilized for training the machine learning models. The cross-validation approach has been used to avoid overfitting problem occurred during training and testing steps. Machine learning models such as Random Forest (RF), Extra Trees (ET), Linear Support Vector Machine (SVM), Non - Linear SVM, and K- Nearest Neighbors (KNN) have been used for classification between ASD and controls. To evaluate the performance of classification, accuracy, precision, recall, and ROC- AUC score values have been considered.