Speech Biomarkers Detect Bipolar Depression Recurrence
Researchers collected 304 voice recordings from 92 clinically diagnosed bipolar disorder patients over one year and developed between-person and within-person machine-learning classifiers to detect moderate-to-severe depression and recurrence. Combined acoustic and linguistic features yielded AUC 0.76 for between-person detection (vs 0.54 demographic) and AUC 0.70 for within-person recurrence detection (vs 0.55), demonstrating feasibility for digital health monitoring.
Key Points
- 1Developed classifiers using 304 recordings from 92 bipolar patients across one year.
- 2Found combined acoustic and linguistic model achieved AUC 0.76, outperforming demographic baseline.
- 3Enables longitudinal within-person monitoring with AUC 0.70, supporting scalable digital health tools.
Scoring Rationale
Solid peer-reviewed longitudinal study with practical models, but moderate novelty and limited generalizability beyond studied cohort.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems
