Objectives:
In a group of patients diagnosed with BD and MDD, an analysis was conducted to evaluate the efficacy of AI algorithms in detecting mental state changes based on physical voice parameters

Methods:
The MoodMon system was developed, including a mobile application for smartphones. In the first stage, the AI was trained using objective data and clinical assessments conducted by psychiatrists, which included 17-item versions of the HDRS (Hamilton Depression Rating Scale) and YMRS (Young Mania Rating Scale) scales, and the CGI (Clinical Global Impression) scale. The second stage was to further refine the AI using individual and population data and generate alerts when subtle changes in mental state were detected. Both stages of the study lasted a total of 944 days.

Results:
Physical voice parameters can serve as biomarkers in affective disorders. The most effective in detecting changes in mental state were 19 specific physical voice parameters. The system showed high performance, with the following sensitivity (true positive rate – TPR) and specificity (true negative rate – TNR) values – for both diagnoses: TPR = 89.5%, TNR = 98.8%; BD: TPR = 89.6%, TNR = 98.9%; MDD: TPR = 89.1%, TNR = 98.5%. MoodMon helped to accurately monitor and predict changes in the mental state of patients with affective disorders.

Conclusions:
The MoodMon system is an objective AI tool that effectively identifies the initial period of mental state changes in affective disorders based on physical voice parameters.