Multi-species numerical validation of an efficient algorithm for auditory brainstem response hearing threshold estimation

Abstract

The Auditory Brainstem Response (ABR) can be used to evaluate hearing sensitivity of animals who are unable to respond to behavioral tasks. However, typical data collection methods are time consuming; decreasing the measurement time may save resources or allow researchers to spend more time on other tasks. Here, an adaptive algorithm is proposed for efficient estimation of ABR thresholds. The algorithm relies on the online update of the predicted hearing threshold from a Gaussian process model as ABR data are collected using iteratively optimized stimuli. To validate the algorithm, ABR threshold estimation is simulated by adaptively sub-sampling pre-collected ABR datasets for which the stimuli were systematically varied in frequency and level. The simulated experiment is performed on 5 datasets of Mouse (2 different datasets), Budgerigar, Gerbil, and Guinea Pig ABRs collected by different laboratories, with a total of 27 ears. The original datasets contain between 68 and 106 stimuli conditions, while the adaptive algorithm is run up to a total of 20 stimuli conditions. The adaptive algorithm ABR threshold estimate is compared against human rater estimates who view the full ABR dataset. The adaptive algorithm threshold matches the human estimates within 10 dB, averaged over frequency, for 19 out of 27 ears. The adaptive procedure is able to provide threshold estimates that are comparable to the human rater estimated thresholds while reducing the measurement time by a factor of 3 to 5. The standard deviation of threshold estimates from successive runs is smaller than the inter-human rater differences, indicating adequate test/retest reliability.