A Bayesian decision analysis approach to assess voice disorder risks by using acoustic features.

Area of Science:

• Occupational Health
• Acoustic Analysis
• Medical Diagnostics

Background:

• Vocal fold nodules are an occupational disease for voice professionals.
• Computer-aided systems offer noninvasive, low-cost voice disease diagnosis.
• Acoustic features from voice recordings can identify vocal fold nodules.

Purpose of the Study:

• To develop a Bayesian decision analysis for classifying vocal fold nodule risk.
• To associate risk levels with appropriate treatments for voice professionals.
• To enable personalized risk assessment and treatment assignment for university lecturers.

Main Methods:

• Extracted acoustic features from voice recordings of healthy individuals and those with vocal fold nodules.
• Applied Bayesian decision analysis to calculate posterior probabilities of vocal fold nodules.
• Incorporated utility functions considering financial costs and recovery probabilities for treatment optimization.

Main Results:

• Successfully classified university lecturers into low, medium, and high vocal fold nodule risk groups.
• Demonstrated the approach's ability to assign individuals to appropriate treatments based on risk.
• Validated the application within the University of Extremadura's Disease Prevention Program.

Conclusions:

• The Bayesian approach provides an effective method for identifying vocal fold nodule risk in voice professionals.
• This system facilitates personalized risk assessment and tailored treatment strategies.
• The methodology is adaptable for various voice-reliant occupations beyond university lecturers.