Wideband acoustic immittance characteristics and machine learning-based diagnostic model for children with large vestibular aqueduct syndrome

Abstract

Objective:This study was to investigate the wideband acoustic immittance（WAI） characteristics of children with large vestibular aqueduct syndrome（LVAS） and to construct a diagnostic model for LVAS based on WAI and machine learning（ML） techniques. Methods:We performed a retrospective analysis of the data from 38 children（76 ears） with LVAS and 44 children（88 ears） with normal hearing. The data included conventional audiological examination, temporal bone CT scan and WAI test. We performed statistical analysis and developed multivariate diagnostic models based on different ML techniques. Results:The two groups were balanced in terms of ear, gender, and age（P>0.05）. The wideband absorbance（WBA） of the LVAS group was significantly lower than that of the control group at 1 000-2 519 Hz, while the WBA of the LVAS group was significantly higher than that of the control group at 4 000-6 349 Hz（P<0.05）. WBA at 5 039 Hz under ambient pressure had a certain diagnostic value（AUC=0.767）. The multivariate diagnostic model had a high diagnostic value（AUC>0.8）, among which the KNN model performed the best（AUC=0.961）. Conclusion:The WAI characteristics of children with LVAS are significantly different from those of normal children. The diagnostic model based on WAI and ML techniques has high accuracy and reliability, and provides new ideas and methods for intelligent diagnosis of LVAS.