Korrelation zwischen Cochlea-Implantat-MAPping und Hörveränderung basierend auf überwachten Lernmodellen des Maschinellen Lernen.

Introduction Cochlear implant (CI) fitting or CI MAPping process takes constant effort to obtain proper settings because the settings depend strongly on the individual. It may be potentially helpful to analyze the correlations between changes in the MAPping and hearing perception by support vector machines (SVM) which are supervised learning models in machine learning (ML).

Methods Datasets of twenty-five patients with unilateral or bilateral CI were investigated (12 male and 13 female) with a mean age of 60 years. Based on the patients’ protocols, a pattern classification was conducted between the change in the MAPping and three frequently occurring hearing perceptions. To find the correlation between them, SVM was used. For the efficient analysis of the pattern, feature extraction based on frequency ranges was conducted, and then the feature importance was calculated to find more relevant features for predicting a specific class.

Results Based on the SVM classifier, a classification model with an accuracy of over 92 % was developed. By using this model, the correlation between the features based on frequency ranges and hearing perception was found. In addition, it was found which feature was more valuable to predict a specific hearing symptom.

Conclusion  The high accuracy of the developed model based on SVM means that the extracted features have a high correlation with hearing perception. This insight into the relationship between them makes the MAPping process more efficient and accurate. However, because of the lack of datasets, further research is necessary to improve the performance of the described ML-trained model.

Publication History

Article published online:
12 May 2023