Analysis of Psychiatric Disorders from EEG Signals Using Machine-Learning Techniques

Abstract

Over the past few years, Psychiatric Disorders (PD) have had a significant impact on global health and their prevalence has been leading towards major adversities like functional disabilities and even suicide. These disorders can be divided into some major and specific categories which show different symptoms and have different remedies accordingly. This study makes an attempt to tackle this problem head on and detect said disorders to allow the patients to take necessary actions before the point of no return. For efficient and trustworthy detection of PD, a Machine Learning (ML) approach has been taken and different algorithms were run on the chosen dataset. The dataset that was used for this study was collected from Neuroimaging (NI), Electroencephalography (EEG), tests which have a variety of distinctive features. It was observed that EEG is a reliable and effective way of collecting brain signals which can later be used in different studies like this one. Judging by the magnitude of the samples taken by the EEG device, it was decided that ML would be a very useful tool in this regard and with good accuracies, an acceptable structure can be created. The goal of this study to make a contribution to application of machine learning algorithms in medical sciences and also to call attention to the capabilities of EEG in the prompt detection of PD. From the findings of this study, it can be observed that very high accuracy was obtained for both the binary and multiclass classifications. The results were tabulated taking samples by using feature selection and feature extraction methods. The highest accuracy for main disorder for multiclass classification was 80.69% and that of the specific disorder was 87.52% both of which used SPARSE PCA feature extraction method. This study makes a solid attempt at addressing this rising issue with a very satisfactory approach and thus makes a fruitful contribution to the medical and data science field for addressing similar adversities in the process