Neural Network-Based Schizophrenia Diagnosis Utilizing Fourier Decomposition

Schizophrenia (SZ) is a mental illness affecting over 20 million individuals globally. It is recognized as a significant concern for our society. This paper introduces a five layer artificial neural network (ANN) classifier to differentiate between SZ patients and healthy controls. The EEG signals are pre-processed with low pass and high pass finite impulse response (FIR) Butterworth filtering, segmentation, and normalization before being decomposed into sub-bands using the Fourier decomposition technique. These sub-bands are then used to extract features like energy, entropy, and Hjorth activity parameter, which are then fed into a simple ANN network to classify the signals. The proposed work achieves an accuracy of 98.97%, sensitivity of 100% and specificity of 98.01%.