Show simple item record

dc.contributor.advisorZouridakis, George
dc.creatorBastarrika, Ainhoa
dc.date.accessioned2013-10-14T20:03:43Z
dc.date.available2013-10-14T20:03:43Z
dc.date.createdAugust 2012
dc.date.issued2012-08
dc.identifier.urihttp://hdl.handle.net/10657/457
dc.description.abstractObjective: In this Master Thesis, we employed auditory evoked responses to identify features from the time domain signals, Granger Causality matrixes and graph theory that provide maximum separation among schizophrenic patients, cocaine addicts and normal controls. Methods: We analyzed data from 12 schizophrenia, 12 cocaine addicts and 12 normal control subjects. Responses were obtained in a paired-stimulus paradigm, in which auditory stimulus S1 is followed by an identical S2. Amplitude and latency of the N100 component were measured from the averaged evoked potential, Granger Causality matrixes were computed and a graph was constructed from there. Amplitude, latency, Granger Causality values and a group of characteristics form the graph were used as features to cluster responses in three groups. Several methods were used for clustering, while their performance was quantified in a 10-fold cross validation approach. Results: We found that the most important features come from Granger Causality matrix values, amplitude and latency coming from the average evoked responses appear as insignificant features for the clustering. Influence of electrode C3 to Pz appear as the most significant feature, separating schizophrenic patients from normal controls and cocaine addicts with a 100% accuracy. In order to separate cocaine addicts from normal controls at least 31 features coming from Granger Causality matrixes were needed. Conclusions: Our results demonstrate that Granger Causality values can accurately separate schizophrenia patients, cocaine addicts and normal controls and suggest that the Pz-C3 region plays a significant role in information processing in human brain. i Significance: The proposed technique may have a significant impact as a clinical tool in the quest for identifying physiological markers of schizophrenia. ii
dc.format.mimetypeapplication/pdf
dc.language.isoeng
dc.subjectAEP
dc.subjectschizophrenia
dc.subjectcocaine
dc.subjectgranger causality
dc.subjectclustering
dc.subject.lcshSchizophrenia
dc.titleUNSUPERVISED CLUSTERING OF SCHIZOPHRENIA PATIENTS, COCAINE USERS, AND NORMAL CONTROLS BASED ON AUDITORY EVOKED POTENTIALS
dc.date.updated2013-10-14T20:03:48Z
dc.type.genreThesis
thesis.degree.nameMaster of Science
thesis.degree.levelMasters
thesis.degree.disciplineEngineering Technology
thesis.degree.grantorUniversity of Houston
thesis.degree.departmentEngineering Technology
dc.contributor.committeeMemberGurkan, Deniz
dc.contributor.committeeMemberYuan, Xiaojing
dc.type.dcmiText
dc.format.digitalOriginborn digital
thesis.degree.majorNetwork Communications
dc.description.departmentEngineering Technology
thesis.degree.collegeCollege of Technology


Files in this item


Thumbnail

This item appears in the following Collection(s)

Show simple item record