A distance-based neurorehabilitation evaluation method using linear SVM and resting-state fMRI
- 1Department of Electronic Engineering, Tsinghua University, China
- 2Department of Rehabilitation Medicine, Beijing Tsinghua Changgung Hospital, China
- 3School of Medicine, Tsinghua University, China
During neurorehabilitation, clinical measurements are widely adopted to evaluate behavioral improvements after treatment. However, it is not able to identify or monitor the change of central nervous system (CNS) of each individual patient. Resting-state functional Magnetic Resonance Imaging (rs-fMRI) has been widely used to investigate brain functions in healthy controls (HCs) and patients with neurological diseases, which could find functional changes following neurorehabilitation. In this paper, a distance-based rehabilitation evaluation method based on rs-fMRI was proposed. Specifically, we posit that in the functional connectivity (FC) space, patients and HCs distribute separately. Linear Support Vector Machines (SVM) were trained on the brain networks to firstly separate patients from HCs. Second, the FC similarity between patients and HCs was measured by the L2-distance of each subject’s feature vector to the separating hyperplane. And finally, statistical analysis of the distance revealed rehabilitation program induced improvements in patients and predicted rehabilitation outcomes. A resting state fMRI dataset with 22 HCs and 18 spinal cord injury (SCI) patients was utilized to validate our method. We built whole-brain networks using 5 atlases to test the robustness of the method and search for features under different node resolution. The classifier successfully separated patients and HCs. Significant improvements in FC after treatment were found for the patients for all 5 atlases using the proposed method, which was consistent with clinical measurements. Furthermore, distance obtained from individual patient’s longitudinal data showed similar trend with each one’s clinical scores, implying the possibility of individual rehabilitation outcome tracking and prediction. Our method not only provides a novel perspective of applying rs-fMRI to neurorehabilitation monitoring, but also proves the potential in individualized rehabilitation prediction.
Keywords: spinal cord injury (SCI), Resting-state fMRI, functional connectivity, Support vector machine, Neurorehabilitation
Received: 12 Jul 2019;
Accepted: 02 Oct 2019.
Copyright: © 2019 Ge, Pan, Wu and Dou. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Dr. Weibei Dou, Tsinghua University, Department of Electronic Engineering, Beijing, 100084, Beijing, China, email@example.com