A research article titled Modeling Markers for Detection of Psychiatric Disorders Using EEG Signals has been co-authored by Dr Lakshman Varanasi, Assistant Professor, Biological Sciences, SIAS; Varun Viswanathan, Visiting Assistant Professor, Psychology, SIAS; Dr Debasish Mishra, Assistant Professor, Data Science and Information Systems, IFMR GSB; and Steven Chris, Teaching Fellow, Data Science and Information Systems, IFMR GSB. The paper has been published in the Journal of Data Science and Intelligent Systems.
Abstract
The diagnosis of mental (psychiatric) disorders is challenging, and there is a lack of consensus on objective diagnostic criteria that are based on definitive signs that accompany the disorder. There is a need, therefore, to develop objective tools for the examination of these disorders. We present here a novel machine learning (ML) approach that accurately identifies disorders. The approach uses electroencephalography (EEG) signals for diagnosis, which are processed to extract novel region based markers that are found to contain key information about the types of disorders. Subsequently, a support vector machine (SVM) classifier is modeled, integrated with sequential feature (marker) selection (SFS), which identifies optimal and compact marker subsets for disorder detection. The proposed system has been validated using a publicly available dataset. The developed model was benchmarked against existing models and was shown to perform superior to the models it was extensively compared with; it demonstrated a 98.33% accuracy in detecting obsessive-compulsive disorder (OCD). Our findings indicate that an accurate psychiatric diagnosis system can be achieved using EEG signals with significantly fewer, and more interpretable markers. This simpler and transparent approach improves the practicality and trustworthiness of AI/ML-driven diagnostic tools, making them more suitable for real-world clinical integration and understanding by medical professionals.