A Novel Automatic Audiometric System Design Based on Machine Learning Methods Using the Brain's Electrical Activity...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Standard audiometry relies on patient responses, which can be subjective.
• Electro-encephalography (EEG) records brain's electrical activity, offering an objective measure.
• Automating hearing tests with machine learning could improve efficiency and accessibility.

Purpose of the Study:

• To develop and validate a machine learning model for autonomous hearing tests using EEG signals.
• To investigate the efficacy of various machine learning algorithms in classifying heard versus unheard sounds based on EEG data.
• To determine the optimal machine learning algorithm for this application.

Main Methods:

• EEG signals were recorded from participants exposed to sounds of varying amplitudes and wavelengths.
• MATLAB was used for stimulus presentation and response recording, while Python was employed for data analysis.
• Machine learning algorithms including Naïve Bayes, LGBM, SVM, decision tree, k-NN, logistic regression, and random forest were applied after EEG data pre-processing and feature extraction (TF-IDF).

Main Results:

• The Light Gradient Boosting Machine (LGBM) algorithm demonstrated the highest performance in classifying audibility.
• The LGBM algorithm achieved an 84% success rate in predicting whether a participant heard a given sound.
• EEG signal analysis using TF-IDF and Count Vectorizer effectively identified relevant features for classification.

Conclusions:

• Machine learning, particularly the LGBM algorithm, shows significant potential for conducting autonomous hearing tests via EEG.
• This approach offers a promising, objective alternative or supplement to traditional audiometry.
• Further research may be needed to integrate this technology into clinical practice, potentially with audiologist oversight.