Classification of Obsessive-Compulsive Disorder by Two-Dimension Convolutional Neural Network

Abstract

Obsessive-compulsive disorder is a mental disorder that a person has uncontrolled, causing the patient to be high anxiety, which affects the daily life of the patient. EEG signal is one method used to diagnose brain diseases. In which EEG is a measure of changes in electrical charges in the brain. In this paper, we propose a twodimension convolution neural network model, a popular way to distinguish between two categories by using images. The time-frequency image was created using the complex Morlet wavelet. Time-frequency images are used as an input of the 2-D convolution neural network model, which uses 2-D convolution to extract the characteristics of the image in each channel of the electrode. The experiment result of testing is 87.5 percent accuracy and 0.461 test loss value in the 2-D CNN model. Therefore, the power of time-frequency in two-dimension convolution neural network model provides a new method for the classification of obsessive-compulsive disorder in the Flanker task.