Font Size: a A A

Depresion EEG Characteristics And Classification Study

Posted on:2015-06-16Degree:MasterType:Thesis
Country:ChinaCandidate:Z Z LiaoFull Text:PDF
GTID:2284330452953264Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
Depression is a affective disorder. The incidence of this disease is high. EEG canreflect different state of brain, including abnormal state. So it is a commonly used toolin clinical psychiatric examination. Our research hope that some different EEGfeatures between normal people and depression can be found. So that we canunderstand the mechanism of depression disease deeply and it is help to diagnosis andto evaluate the therapeutic effect.Our research samples from3groups are the normal people, the unmedicateddepression and the medicated depression. The normal group is14persons. Theunmedicated group is11people. And the medicated group is11people too. Subjectsare got their EEG without any stimulate when they are awake. The EEG lasts8minutes.We research deeply in absolute power, relative power ratio, left-right asymmetryand stable process difference in each band in each group. We find some importantresults. One is that alpha1, alpha2bands of absolute power; theta band of left-rightasymmetry; alpha2band of stable process difference are the nature dependentcharacteristic. The other is that delta band of absolute power; alpha1, theta band ofrelative power ratio; alpha1, alpha2, beta1, delta band of left-right asymmetry; theta,delta band of stable process difference are the state dependent characteristic. Beta1band of absolute power, beta1band of relative power ratio can reflect side effect ofantidepressant.We also do automatic EEG classification in our research. The algorithm what weused are C4.5decision tree, naive bayes and K nearest neighbor. The characteristic weuse are absolute power, the relative power ratio, left-right asymmetry and stableprocess difference. In each characteristic, we research7bands and all bands.Classification results are in accordance with T test in many situations. The bestclassification correct rate can arrive at70.00%using C4.5decision tree, and it canarrive83.33%using naive bayes, it can arrive70.37%using K nearest neighbor. Inaddition, the nature dependent characteristic and the state dependent characteristic areexamined.
Keywords/Search Tags:depression, power spectrum, EEG characteristic, EEG pattern identify
PDF Full Text Request
Related items