Machine Learning Approaches for Obsessive Compulsive Disorder Detection

Abstract Obsessive-Compulsive Disorder (OCD) is a psychiatric illness that produces significant psychological distress in patients. Individuals with OCD have recurring unwanted thoughts or sensations which make them obsessed with something and feel to do something repetitively as a compulsion. In general detection of OCD is performed by symptoms analysis. However, the symptoms are significantly visible at a later stage. Even individuals with OCD have less faith in the analysis of the symptoms as long as it is not affecting their life negatively. As a result, they start their treatment at a later stage and the treatment process becomes longer. However, it is observed that if the detection is performed through laboratory analysis through some biomarkers then the patients have more faith in the detection process and can start their treatment well in advance. Therefore laboratory detection of OCD can play a vital role in OCD treatment effectiveness. Most of the laboratory detection process proposed in the literature uses Machine Learning on related biomarkers. However, the prediction accuracy rate is not enough. This research aims to analyze the approaches to pediatric OCD based on machine learning using neuroimaging biomarkers and oxidative stress biomarkers. The challenges in OCD detection and prediction using neuroimaging biomarkers, oxidative stress biomarkers, and Machine Learning models have been described. Further, it analyzes the performance of different machine learning models that were used for OCD detection and highlights the research gap to improve prediction accuracy.