Differentiation of Auditory Event-Related Potentials from Normal Subjects Musicians and Patients with Schizophrenia by Two-Stage Hierarchical Classification Strategy

  • 孫 銘鍵

Student thesis: Master's Thesis

Abstract

Schizophrenia is a severe and chronic brain disorder that affects about 1% population worldwide and it typically occurs in young adulthood Common symptoms include hallucinations delusions social withdrawal reduced emotional expression etc Currently diagnosis of schizophrenia is based on observing the patient's behavior and reported experiences for 6 months or more (Diagnostic and Statistical Manual of Mental Disorders DSM) Past studies showed that auditory perception had long been used to test theories for neural system or develop assessment system for illness such as neuroplasticity in musician’s brain and auditory processing deficits in schizophrenia patients In previous studies we found that difference in auditory event-related potentials (AEP) existed among normal subjects musicians and patients with schizophrenia Therefore the aim of this study is to classify normal subjects musicians and patients with schizophrenia by analyzing objective electrophysiological information This study constructs a classification strategy which can not only develop an effective and objective classification system to help diagnosis of schizophrenia but also distinguish musicians who have specialized auditory perception from normal people In this thesis the auditory event-related potentials experiments were recorded from 12 well-trained musicians 12 medicated schizophrenia patients and 12 normal subjects The stimuli was presented randomly by three kinds of chord stimuli (major diminish and atonal) and two kinds of interval stimuli (perfect fifth and tritone) by observing the statistical results of AEPs in N1 and P2 amplitude of each stimuli type in groups and selected which had statistical difference as features and then combined with two-stage strategy to select best feature combinations for classification Results show that when firstly separating schizophrenia patients from the others in stage one and then separating musicians and normal subjects in stage two the best overall accuracy can reach 91 67% The accuracy for schizophrenia classification can achieve 97 22%; the accuracy for musician classification can achieve 91 67% Hence this thesis proposed a classification strategy which can not only correctly classify schizophrenia patients but also correctly classify musicians Finally this study classified three groups who had distinct levels of auditory perception by using discriminating features and two-stage hierarchical strategy Furthermore this also provided a high accuracy effective and objective classification system for clinical utility to help diagnosis of schizophrenia
Date of Award2014 Aug 27
Original languageEnglish
SupervisorSheng-Fu Liang (Supervisor)

Cite this

Differentiation of Auditory Event-Related Potentials from Normal Subjects Musicians and Patients with Schizophrenia by Two-Stage Hierarchical Classification Strategy
銘鍵, 孫. (Author). 2014 Aug 27

Student thesis: Master's Thesis