DTI Based Diagnostic Prediction of a Disease via Pattern Classification

摘要: The paper presents a method of creating abnormality classifiers learned from Diffusion Tensor Imaging (DTI) data population patients and controls. score produced by the classifier can be used to aid in diagnosis as it quantifies degree pathology. Using anatomically meaningful features computed DTI we train non-linear support vector machine (SVM) pattern classifier. begins with high dimensional elastic registration DT images followed feature extraction step that involves concatenating average anisotropy diffusivity values regions. Feature selection is performed via mutual information based technique sequential elimination features. A SVM then constructed training on selected assigns each test subject probabilistic indicates extent In this study, were created for two populations; one consisting schizophrenia (SCZ) other individuals autism spectrum disorder (ASD). clear distinction between SCZ controls was achieved 90.62% accuracy while ASD, 89.58% classification obtained. scores clearly separate groups prospect using diagnostic prognostic marker.