Depression Detection Using Blood Cortisol Levels with Machine Learning Algorithms

Depression is a pervasive mental health disorder, and timely and accurate diagnosis is critical for effective treatment. This research explores the feasibility of using blood cortisol levels as a biomarker for detecting depression. Through the utilization of machine learning algorithms, our objective is to construct a predictive model capable of categorizing individuals as either depressed or non-depressed based on their blood cortisol levels. A diverse and well-defined group of participants underwent standardized depression assessments, accompanied by the analysis of their blood samples to determine cortisol levels. Machine learning techniques, including Random Forest, Support Vector Machines, and Logistic Regression, among others, were employed to develop and validate the depression detection model. In this fictitious scenario, the test set performance metrics reveal that the SVM model achieved an accuracy of 0.85, precision of 0.82, recall of 0.87, and F1-score of 0.84. The GMM model showed slightly lower metrics, with an F1-score of 0.73, accuracy of 0.68, precision of 0.79, and recall of 0.79. Notably, the CNN model outperformed the others, boasting a remarkable 0.92 F1-score, 0.92 accuracy, 0.91 precision, and 0.93 recall. These results underscore the potential of using machine learning and blood cortisol levels as a reliable and objective tool for early depression detection, thereby enhancing the overall quality of mental health care outcomes.