Machine learning-based longitudinal prediction for GJB2-related sensorineural hearing loss.

Area of Science:

• Genetics
• Audiology
• Machine Learning

Background:

• Recessive GJB2 variants are the primary genetic cause of hearing loss.
• These variants can lead to progressive sensorineural hearing loss (SNHL).

Purpose of the Study:

• To develop a machine learning model for predicting GJB2-related SNHL progression.
• To enable personalized medical planning and timely intervention.

Main Methods:

• A nationwide cohort of 449 patients with biallelic GJB2 variants was analyzed.
• Machine learning models, including Long Short-Term Memory (LSTM), were trained and validated.
• Model performance was assessed using Mean Absolute Error (MAE).

Main Results:

• Hearing loss progression was observed in all models, averaging 0.61 dB HL/year.
• The LSTM model achieved the best performance with an MAE of 4.34 dB HL.
• The model demonstrated acceptable accuracy for predicting hearing loss up to 4 years.

Conclusions:

• A prognostic machine learning model for GJB2-related SNHL was successfully developed.
• The model facilitates individualized medical plans and optimal follow-up intervals.
• This approach supports personalized management of genetic hearing loss.