Automatic ABR detection at near-threshold intensities combining template-based approach and energy analysis

Abstract

This paper presents an algorithm for detecting auditory brainstem responses elicited by click stimuli at low intensities. Feature extraction is focused on peak V detection including analysis of the instantaneous energy and comparison with a template constructed from a reference dataset. All detection methods were validated and adjusted using ROC curve analysis. The algorithm was implemented in MatLab and evaluated in 135 ABR recordings. The combination of sensitivity, specificity and ROC area values exceeding 95 % with a 35 ms running time validates this approach for fast and accurate ABR detection.