Diagnosis of schizophrenia from R-fMRI data using Ripplet transform and OLPP

• Published in: Multimedia tools and applications Vol. 79; no. 31-32; pp. 23401 - 23423
• Main Authors: Sartipi, Shadi, Kalbkhani, Hashem, Shayesteh, Mahrokh G.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.08.2020; Springer Nature B.V
• Subjects: Algorithms; Brain; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Diagnosis; Feature extraction; Independent component analysis; Magnetic resonance imaging; Multimedia Information Systems; Schizophrenia; Special Purpose and Application-Based Systems; Support vector machines; Computer-aided diagnosis (CAD); Schizophrenia; Orthogonal locality preserving projection (OLPP); Functional magnetic resonance imaging (fMRI); Ripplet transform
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-020-09122-y

Schizophrenia is a severe brain disease that influences the behaviour and thought of person. These effects may fail in achieving the expected levels of interpersonal, academic, or occupational functioning. Although the underlying mechanism is not yet clear, the early detection of schizophrenia is an attractive and challenging research area. There are differences in brain connections of patients and healthy people. This study presents a new computer-aided diagnosis (CAD) method to diagnose schizophrenia (SZ) patients from normal control (NC) people by using the rest-state functional magnetic resonance imaging (R-fMRI) data. fMRI data has a huge dimension, and extracting efficient features is still an open challenge for a schizophrenia diagnosis. In the proposed method, at first orthogonal locality preserving projection (OLPP) is used to reduce the number of time points in R-fMRI scans. Then, an independent component analysis (ICA) algorithm is employed to estimate the independent components (ICs). Next, orthogonal Ripplet-II transform is applied to each IC to extract features. Afterward, a two-sample T-test is implemented on the extracted features to find the most discriminative features. Then, the number of selected features is reduced by applying OLPP. Finally, a test subject is classified into SZ or NC using a linear support vector machine (SVM) classifier. The proposed method is evaluated on the NAMIC and COBRE databases. The results demonstrate that the introduced method significantly outperforms previously presented methods.