An Intelligent Psychiatric Recommendation System for Detecting Mental Disorders

Abstract

The inadequacy of the number of specialists and the resultant heavy workloads impede diagnostic efforts, making it very difficult to receive appropriate medical services and manage the treatment process. Such problems underscore the need for auxiliary systems to help experts in making diagnoses, saving both labor and time. For this reason, we propose a new intelligent psychiatric recommendation system with the Comprehensive Psychiatric Differential Diagnosis Test (CPDDT), which we created to screen and differentiate among psychiatric diagnoses. To guide experts in using the system, we included axis one and axis two diagnosis groups, which, respectively, refer to clinical and personality disorders in the DSM-4. The goal was to measure areas affecting the course of an illness and the treatment plan developed by a specialist, including functionality, memory, and suicidal thoughts. The CPDDT can detect 48 different diagnostic groups from the answers to 319 questions. The system was subjected to an online test of 676 users via a web system developed by DNB Analytics. Psychiatrists evaluated the results in a clinical setting. The test results were then evaluated by the evolutionary simulation annealing LASSO logistic regression model. After determining the importance of each question on the scale, the algorithm eliminated the questions with the least impact and the test was reduced to 147 questions, producing a .93 level of accuracy. In addition, the algorithm found the probability of each patient suffering from a disorder. In summary, the new machine-learning-based CPDDT was finalized to include 147 questions; the algorithm is presented here as a useful suggestion system for experts engaging in the diagnostic process.