SVM Identification of Early Vascular Dementia Patients

Chao Wang1, Jin Xu1, Songzhen Zhao1

  • 1Xi'an Jiaotong University

Details

11:30 - 13:30 | Fri 26 May | Emerald III, Rose, Narcissus & Jasmine | FrPS1T1.40

Session: Poster I

Abstract

Much effort has been made to find neuroimaging markers for diagnosis of vascular dementia (VaD). In this paper, we combined EEG and support vector machine (SVM) to differentiate early VaD patients from controls. Interregional directed connectivity and resulting topological parameters were extracted and entered into classification respectively. It was found that VaD patients can be correctly separated from controls with high accuracy. A large proportion of discriminant features were associated with parietal and temporal regions. This was highly consistent with those obtained through traditional statistical tests. The study may facilitate automatic VaD diagnosis at an early stage in future clinical practice.