Explainable machine learning model for predicting hearing recovery in unilateral sudden sensorineural hearing loss

Area of Science:

• Otolaryngology
• Medical Informatics
• Machine Learning

Background:

• Sudden Sensorineural Hearing Loss (SSNHL) prognosis is variable and influenced by numerous factors.
• Accurate prediction of SSNHL outcomes is crucial for effective patient management and rehabilitation.

Purpose of the Study:

• To develop an explainable Machine Learning (ML) model for predicting SSNHL prognosis.
• To identify easily accessible features that significantly impact hearing recovery in SSNHL patients.

Main Methods:

• A bi-center retrospective study involving 534 SSNHL patients.
• Feature selection using univariate analysis and LASSO regression.
• Development and evaluation of five ML models using AUC, accuracy, specificity, sensitivity, and F1 scores, with SHAP for interpretability.

Main Results:

• The Random Forest (RF) model achieved an AUC of 0.998, with high accuracy (0.981), sensitivity (0.963), specificity (0.989), and F1 score (0.967).
• Key predictive features identified include degree of hearing loss, audiogram type, age, Mean Corpuscular Volume (MCV), onset to treatment interval, and serum Albumin (ALB).

Conclusions:

• An explainable ML model demonstrates excellent performance in predicting SSNHL hearing outcomes.
• The SHAP method provides valuable insights into feature contributions, assisting clinicians in understanding hearing recovery drivers and guiding aural rehabilitation.