Using image processing technology combined with decision tree algorithm in laryngeal video stroboscope automatic...

Area of Science:

• Otolaryngology
• Medical image analysis
• Computational pathology

Background:

• Laryngeal video stroboscopy is crucial for diagnosing vocal fold pathologies.
• Objective physiological data extraction from clinical videos remains challenging.
• Existing diagnostic methods may lack quantitative analysis.

Purpose of the Study:

• To develop an automated system for vocal fold disease identification using laryngeal videos.
• To extract quantitative physiological parameters from dynamic vocal fold images.
• To classify vocal fold conditions including normal, paralysis, and nodules.

Main Methods:

• Utilized clinical laryngeal video stroboscope recordings as primary data.
• Implemented image processing to automatically identify the largest glottal area.
• Employed a decision tree algorithm for vocal fold disease classification.

Main Results:

• The automated system achieved an initial identification rate of 92.6% for vocal fold diseases.
• An image recognition improvement procedure enhanced the identification rate to 98.7%.
• The system successfully differentiated between normal vocal folds, vocal paralysis, and vocal nodules.

Conclusions:

• The developed automatic vocal fold disease identification system demonstrates high accuracy.
• Image processing and decision tree classification offer a viable approach for objective laryngeal diagnostics.
• The system holds significant value for clinical practice in otolaryngology.