Specificity Study Of EEG Signals For Poststroke Depression

Poststroke depression (PSD), referred to as emotional disorders with mainsymptoms of depression and interest drops, was among the most common emotionaldisorders afflicting stroke sufferers, which impacts the rehabilitation progress andfunctional occurrence seriously. At present, PSD remains no uniform standard for theclinical diagnosis and lack of professional equipment or specific physiologicalindexes to depression assessment, which lead to high rate of misdiagnosis and misseddiagnosis.To explore the feasibility of EEG signals using for the diagnosis of PSD, thisthesis acquired resting state EEG signals of three groups (healthy subjects, poststrokenon-depression subjects and poststroke depression subjects) for analysis. Non-lineardynamic characteristics (Lemp-Ziv Complexity, LZC and Sample Entropy, SampEn)and brain network features based on PDC (Partial Directed Coherence) were extractedfor statistical analysis and network topology analysis respectively with the purpose ofexploring specific characteristic changing of EEG signals of PSD subjects.The results of non-linear dynamic analysis showed that the complexity of strokesubjects’ EEG signals was lower than healthy subjects in most of brain regions andthe reduction of complexity in PSD subjects’ EEG signals was more significant infrontal lobe, temporal lobe and occipital lobe. The correlation study found that theseverity of poststroke depression was correlated with the complexity of EEG signalsin frontal region significantly (P<0.05), indicating that the neurons electrical activityof PSD subjects was more ordered and simple than the other two group subjects.Another correlation study found that the time post stroke was correlated with thecomplexity of α-band EEG signals of most brain regions significantly (P<0.05),implicating that neurons electrical activity of α rhythm was associated withrehabilitation of stroke patients.With the above non-linear characteristics to be feature space for classification,the highest classification accuracy could be up to85%between poststrokenon-depression subjects and poststroke depression subjects, indicating that the abovenon-linear dynamic characteristic had a good degree of separability and could beexpected to become effective diagnostic index to PSD. The results of brain networkanalysis based on PDC showed that the cluster coefficients of brain network reduces from healthy controls to poststroke non-depression subjects, and to poststrokedepression subjects. Besides, the betweenness centrality was increasingly dispersed,which indicating that the “core node” of PSD subjects has been transferred comparedwith postroke non-depression subjects and healthy controls, and the status of corenode” has been declined, suggesting that the mood regulatory pathway has been “lossof connection” with the decline of network operation efficiency. Achievements of thispaper revealed the electrophysiological mechanisms of PSD, and could providetechnical support for the clinical diagnosis of PSD.