Depression Severity Classification from Speech Emotion.

Area of Science:

• Neuroscience
• Computational Psychiatry
• Machine Learning in Healthcare

Background:

• Major Depressive Disorder (MDD) is a prevalent psychiatric condition.
• Objective audio analysis offers a potential tool for monitoring depression severity during treatment.
• Deep Brain Stimulation (DBS) is an intervention for severe MDD, requiring effective patient monitoring.

Purpose of the Study:

• To develop an automated method for classifying depression severity in MDD patients undergoing DBS treatment.
• To leverage emotion-based audio features for predicting mood states.
• To improve clinical decision-making during DBS therapy for MDD.

Main Methods:

• Utilized audio recordings from MDD patients.
• Employed a Deep Neural Network (DNN) trained on an auxiliary emotion dataset.
• Extracted low-dimensional representations from audio for classification.
• Developed a predictive model based on emotion-derived audio features.

Main Results:

• The proposed audio analysis approach achieved an Area Under the Curve (AUC) of 0.80.
• Effectively classified between depressed and improved phases of DBS treatment.
• Demonstrated superior performance compared to alternative methods.

Conclusions:

• Audio analysis, particularly using DNNs and emotion-based features, is a promising method for objective depression severity classification.
• This approach can support clinical management of MDD patients during DBS therapy.
• Further research can refine this technique for broader clinical application.