Demographic and Symptomatic Features of Voice Disorders and Their Potential Application in Classification Using...

Area of Science:

• Otolaryngology
• Medical Informatics
• Computational Biology

Background:

• Voice disorder screening traditionally relies on symptom questionnaires.
• This study explores using demographic and symptomatic features for computerized voice disorder classification.
• Investigates the potential of machine learning in differentiating specific laryngeal conditions.

Purpose of the Study:

• To determine if demographic and symptomatic features can differentiate between glottic neoplasm, phonotraumatic lesions, and unilateral vocal palsy.
• To evaluate the effectiveness of various machine learning algorithms for voice disorder classification.
• To identify key features contributing to accurate classification of voice disorders.

Main Methods:

• Recruited 100 patients with glottic neoplasm, 508 with phonotraumatic lesions, and 153 with unilateral vocal palsy.
• Applied statistical analyses to identify significant differences in demographic and symptomatic variables.
• Utilized machine learning algorithms (decision tree, LDA, KNN, SVM, ANN) for classification.

Main Results:

• Demographic features were more effective for neoplastic and phonotraumatic lesions; symptoms were better for vocal palsy.
• The artificial neural network (ANN) achieved the highest classification accuracy (83 ± 1.58%).
• Decision tree analysis identified sex, age, smoking status, sudden dysphonia onset, and VHI scores as significant classification features.

Conclusions:

• Significant differences in demographic and symptomatic features exist between glottic neoplasm, phonotraumatic lesions, and vocal palsy.
• These distinct features enable automatic classification of voice disorders using machine learning.
• Machine learning, particularly ANNs, shows promise for objective and accurate voice disorder diagnosis.