Analysis of bipolar disorder using multimodality data

Abstract

Medical Imaging is a challenging task to radiologists for diagnosis. newlineThis may, sometimes leads to wrong treatment because of human error. newlineTherefore, radiologists may use the Computer Aided Diagnosis for correct newlinediagnose. newlinePsychiatric diseases are more common in this century due to the newlinechange of lifestyle and exposure in social media. The most common mental newlineillness which affects both young and old persons are bipolar disorder and newlineschizophrenia that affects the human behavior drastically. We have chosen newlinefMRI and T1 MRI imaging modalities for this study. newlineIn the first study, extracted structural property of T1 MRI image newlinebased on the proposed two invariant features: 3-Dimensional (3D) Scale newlineInvariant Feature Transform and 3D Speeded-Up Robust Feature vectors to newlinediagnose Bipolar Disorder. Kernel based Principal Component Analysis is newlineused to project the feature vectors and given the data to Random Forest. The newlineresults revealed that the method has high potential to identify the bipolar newlinedisorder than earlier works, and an average accuracy of 77.77% is reached. newlineThis research reveals that neuroimaging studies will help to differentiate newlinebipolar disorder from healthy controls. newlineIn the second study, used 3D texture features which assesses newlineproperties like smoothness, coarseness, and regularity of each surface as well newlineas intensity-based features also extracted and best features are selected using newlineGenetic algorithm. newline