Photo Credit: Andrey Popov

A machine learning model based on pure-tone audiometry features can diagnose Meniere disease (MD) and predict endolymphatic hydrops (EH), according to findings published in Otolaryngology-Head and Neck Surgery. Researchers collected gadolinium-enhanced MRI sequences and pure-tone audiometry data. Using air conduction thresholds from pure-tone audiometry, the researchers engineered basic and multiple analytical features. Five classical machine learning models were trained to diagnose MD using engineered features, and the top-performing models were selected to predict EH. The winning light gradient boosting (LGB) model demonstrated remarkable performance for diagnosing MD, with an accuracy of 87%, sensitivity and specificity of 83% and 90%, and an area under the receiver operating characteristic curve of 0.95, comparing favorably with experienced clinicians. The LGB model had 78% accuracy for EH prediction and outperformed the other models. The specific features essential for MD diagnosis and EH prediction include standard deviation and mean of the whole-frequency hearing, audiogram peak, and hearing at low frequencies.