Automated Speech Analysis in Bipolar Disorder: The CALIBER Study Protocol and Preliminary Results

Area of Science:

• Psychiatry
• Computational Linguistics
• Speech Technology

Background:

• Bipolar disorder (BD) is characterized by mood and energy fluctuations.
• Speech patterns in BD patients reflect these shifts, but current assessment is subjective.
• Natural language processing (NLP) presents an opportunity for objective speech analysis in BD.

Purpose of the Study:

• Correlate speech features with manic and depressive symptom severity in bipolar disorder.
• Develop predictive models for diagnosis and treatment outcomes in BD.
• Identify the most relevant speech features and tasks for BD analysis.

Main Methods:

• Longitudinal audio recordings of BD patients (euthymic, manic, depressed, post-response).
• Speech feature extraction (acoustics, content, formal aspects, emotionality) after diarization and transcription.
• Statistical analyses including correlation with clinical scales and lasso logistic regression for predictive modeling.

Main Results:

• Data collected from 76 BD patients across different mood states.
• Preliminary analysis shows distinct speech patterns correlating with symptom severity (YMRS, HDRS-17 scores).
• Ongoing statistical analyses will explore correlations and build predictive models.

Conclusions:

• Automated speech analysis can provide objective biomarkers for psychopathological changes in BD.
• This approach may enhance diagnosis, monitoring, and prediction of treatment response.
• Standardized protocols are vital for a global speech cohort to advance BD research.