THE 2nd INTENATIONAL PEDIATRIC AUDIOLOGY CONGRESS JOINT WITH 15th INTERNATIONALSYMPOZIUM EURO-CIU, 4–6 APRIL 2025, ISTANBUL, TÜRKIYE

Introduction: Auditory Brainstem Response (ABR) test is an important electrophysiological assessment used together with behavioral tests to diagnose hearing loss. Aim: The aim of this study is to estimate the intensity threshold and latency of the V wave in ABR tests performed in children using machine learning methods. In this way, it is aimed to save time during the ABR test, contribute to early diagnosis in cases where the test cannot be performed due to sleep problems in children with special needs, and obtain more objective and reliable results by preventing late diagnosis problems. Material and methods: In this study, a dataset consisting of 88 ABR waveforms consisting of 32 hearing-impaired and normal-hearing individuals from the pediatric group between the ages of 0–8 was analyzed using machine learning techniques. The V wave latency values of the ABR test were estimated using the deep learning model Artificial Neural Networks (ANN). The model was trained using the Levenberg-Marquardt technique for optimization using the MATLAB program. Results: R2 values were used to evaluate the performance of the model. The corresponding R2 values of the training, validation and test sets were determined as 0.99, 0.97 and 0.92, respectively. The Levenberg–Marquardt algorithm was used to optimize the training and test parameters for data evaluation, aiming to estimate wave V features with maximum precision. A classification accuracy of 92.9% was achieved for ABR waveforms matched with the trained machine learning model. Conclusions: The preliminary results obtained demonstrate the potential of machine learning models to objectify and increase the accuracy of ABR interpretation. The results of this study are expected to significantly increase the diagnostic eff iciency in pediatric audiology. The potential to reduce the time and level of expertise required for ABR assessments is expected to support a faster and more comfortable assessment process for patients and audiologists. The use of machine learning models has the potential to reduce the workload of audiologists in a clinical setting by enabling faster and more accurate ABR tests.