Machine studying mannequin can diagnose Meniere illness

A machine studying mannequin primarily based on pure-tone audiometry options can diagnose Meniere illness (MD) and predict endolymphatic hydrops (EH), in keeping with a examine revealed on-line Aug. 28 in Otolaryngology-Head and Neck Surgical procedure.

Xu Liu, M.D., from Fudan College in Shanghai, and colleagues collected gadolinium-enhanced magnetic resonance imaging sequences and pure-tone audiometry knowledge in a retrospective examine. Primarily based on the air conduction thresholds of pure-tone audiometry, primary and a number of analytical options had been engineered. The engineered options had been used to coach 5 classical machine studying fashions to diagnose MD. The fashions with wonderful efficiency had been chosen to foretell EH.

The researchers discovered that the successful gentle gradient boosting (LGB) machine studying mannequin demonstrated outstanding efficiency for analysis of MD, attaining accuracy of 87%, sensitivity and specificity of 83 and 90%, and an space underneath the receiver working attribute curve of 0.95, evaluating favorably with skilled clinicians.

The LGB mannequin had 78% accuracy for EH prediction and outperformed the opposite three fashions. Particular pure-tone audiometry options which can be important for each MD analysis and EH prediction embody normal deviation and imply of the whole-frequency listening to, audiogram peak, and listening to at low frequencies (notably at 250 Hz).

“The examine confirmed promising diagnostic capabilities, indicating the potential of machine studying primarily based on pure‐tone audiometry outcomes as an MD screening methodology,” the authors write.

© 2024 HealthDay. All rights reserved.