An Artificial Neural Network for Depression Screening and Questionnaire Refinement in Undergraduate Students

Depression is unfortunately a very common illness, affecting over 264 million people worldwide, which in extreme cases can lead to suicide. While there are treatments for mental disorders, including depression, many people do not receive adequate treatment or even psychological attention due to lack or resources, social stigmas, inaccurate assessments, and lack of trained mental health professionals. In this paper, a system for screening depression using an artificial neural network is proposed. A true/false questionnaire consisting of 117 items was designed by a medical health professional based on the Diagnostic and Statistical Manual of Mental Disorders (DSM-5). This questionnaire was applied to 157 undergraduate students, and their answers served to train the neural network to fit the related physical symptoms of depression, showing encouraging results in comparison with other machine learning techniques. Posterior principal component analysis and genetic algorithms-based approaches were used to propose methodologies to refine questionnaires, identifying some items which could prove to be more relevant than others, increasing the quality of the results in future survey-related applications.