PsyNet: A robust Deep Learning-based framework for Detecting Psychiatric Disorders using EEG signals

In the modern world, many people suffer from various psychiatric disorders. Diagnosing psychiatric disorders early increases the chances of full recovery for patients and prevents their health from deteriorating. However, detecting psychiatric disorders early is often difficult. The goal of this paper is to design a novel Artificial Intelligence (AI)-powered methodology for detecting different psychiatric disorders by analyzing the EEG (electroencephalography) signals of humans. A deep learning-based classifier model named PsyNet is designed for this purpose. The designed PsyNet model consists of a stacked autoencoder network and an attention mechanism-based proposed PLSTM (Potent Long Short-Term Memory) classifier model. The designed PsyNet model is capable of detecting six different psychiatric disorders, such as addiction disorder, trauma and stress-related disorder, mood disorder, schizophrenia, anxiety-related disorder, and obsessive-compulsive disorder, using the extracted EEG features from the EEG signals of the prefrontal lobe, frontal lobe, parietal lobe, motor cortex, temporal lobe, and occipital lobe. The classification accuracy of the trained PsyNet model is 95.97%. The novelty of this research lies in i) proposing a new architecture of LSTM named PLSTM and using it as a classifier, and ii) incorporating the attention mechanism in the PLSTM classifier network of the PsyNet model to improve its prediction accuracy.