Mood Disorder Severity and Subtype Classification Using Multimodal Deep Neural Network Models.

Area of Science:

• Biomedical Engineering
• Computational Psychiatry
• Neuroscience

Background:

• Mood disorder diagnosis relies on subjective assessments and limited tools.
• Heart Rate Variability (HRV) data offers insights into autonomic nervous system balance.
• Existing statistical methods for HRV analysis show limitations in mood disorder classification.

Purpose of the Study:

• To develop and evaluate novel algorithms for mood disorder subtype and severity classification.
• To compare the performance of deep learning models against traditional statistical analyses.
• To explore the utility of multimodal HRV data analysis for psychiatric diagnosis.

Main Methods:

• Collected heart-related data, including time and frequency domain HRV variables.
• Developed three mood disorder classification algorithms using multimodal deep neural network analysis.
• Compared deep learning model performance with established statistical analysis methods.

Main Results:

• Deep learning analysis demonstrated improved classification accuracy for major depressive disorder (MDD), anxiety disorder (AD), and bipolar disorder (BD) by 0.118, 0.231, and 0.125, respectively.
• Multimodal analysis of HRV data enhanced the identification of mood disorder subtypes and severity.
• The proposed deep learning approach outperformed traditional statistical methods in classification tasks.

Conclusions:

• Deep learning analysis of HRV biomarker data shows promise as a primary tool for mental health diagnosis.
• This approach can aid psychiatrists in objectively diagnosing mood disorders and assessing current mood status.
• Multimodal HRV analysis offers a novel, data-driven method for improving psychiatric diagnostic accuracy.