A Hybrid artificial intelligence method to classify trichotillomania and obsessive compulsive disorder
Erguzel, Turker Tekin
Hizli Sayar, Gokben
MetadataShow full item record
Classification of psychiatric disorders is becoming one of the major focuses of research using artificial intelligence approach. The combination of feature selection and classification methods generates satisfactory outcomes using biological biomarkers. The use of quantitative electroencephalography (EEG) cordance has enhanced the clinical utility of the EEG in psychiatric and neurological subjects. Trichotillomania (TTM), a kind of body focused repetitive behavior, is defined as a disorder characterized by repetitive hair pulling that results in noticeable hair loss. Phenomenological observations underline similarities between hair-pulling behaviors and compulsions seen in obsessive-compulsive disorder (OCD). Despite the recognized similarities between OCD and TTM, there is evidence of important differences between these two disorders. In order to dichotomize the subjects of each disorder, artificial intelligence approach was employed using quantitative EEG (QEEG) cordance values with 19 electrodes from 10 brain regions (prefrontal, frontocentral, central, left temporal, right temporal, left parietal, occipital, midline, left frontal and right frontal) in 4 frequency bands (delta, theta, alpha and beta). Machine learning methods, artificial neural networks (ANN), support vector machine (SVM), k-nearest neighbor (k-NN) and Naïve Bayes (NB), were used in order to classify 39 TTM and 40 OCD patients. SVM, with its relatively better performance, was then combined with an improved ant colony optimization (IACO) approach in order to select more informative features with less iterations. The noteworthy performance of the hybrid approach underline that it is possible to discriminate OCD and TTM subjects with 81.04% overall accuracy.