On the Use of Machine Learning for Classifying Auditory Brainstem Responses: A Scoping Review

Abstract

Recent advances in machine learning have led to a surge of interest in classification of the auditory brainstem response. By conducting a search in the PubMed, Google Scholar, SpringerLink, ScienceDirect, and Scopus databases, it was possible to identify twelve studies that explored the use of machine learning to classify the auditory brainstem response as a complementary and objective method to (a) help clinicians better diagnose hearing impairment by discerning between healthy and pathological auditory brainstem response waveforms, (b) present a neural marker for potential applications in hearing aid tuning, and (c) provide a biometric marker for discriminating between subjects. A comparison between the studies presented in this review is not possible as they used different test subjects, group sizes, and stimuli, and evaluated ABR differently. Instead, the result of these studies will be presented and their limitations as well as their potential applications will be discussed. Overall, the findings of these studies suggest that ABR classification using machine learning is a promising tool for assessing patients with hearing loss, optimizing technologies for tuning hearing aids, and discriminating between subjects.

Application of Signal Classifiers in Auditory Evoked Potentials for the Detection of Pathological Patients

“…There is some background about the application of classifications methodologies using evoked potential records, see for example [19] [22] and [2]. Evaluation of the progression of the Multiple Sclerosis Disease has been studied in [39] and distinguishing age from infants can be seen in the work presented in [29].…”

Multi-Transfer Learning Techniques for Detecting Auditory Brainstem Response

“…Artificial intelligence (AI) refers to the use of computers to automate complex tasks generally performed by humans [8,9]. Machine learning (ML) is a type of AI that is a powerful method of data analysis that is based on the concepts of learning and discovering data patterns rather than being programmed [10].…”

“…After preparing a patient for the test and performing the ABR test, an expert and trained audiologist would analyze and interpret ABR waveforms using the latency, inter-peak latency, and amplitude of ABR peaks [7]. As interpreting the ABR results needs lots of Specialty and skills, less expert clinicians may misinterpret the results and misdiagnose [8]. Additionally, the presence of distorted ABR waveform morphology in some specific disorders such as Central Auditory Processing Disorders (CAPD) further raises the risk of misdiagnosis [2].…”

Translational Applications of Machine Learning in Auditory Electrophysiology

“…For example, do we want an ML algorithm to learn from old clinical ABR datasets and tell us whether a new ABR is normal or abnormal? 6,7 Do we want to use speech evoked AEPs to objectively know whether a hearing aid or cochlear implant user can discriminate between speech sounds? 8 Do we want to know whether auditory training has positively impacted the function of the auditory nervous system?…”

“…It is likely that classical supervised ML approaches will be incorporated into commercially available AEP equipment in the future, allowing users to leverage their own clinical or research databases to train their system to classify or predict new data. Implementation of such a feature could automate aspects of diagnostic testing (e.g., ABR threshold estimation 6,7 ) and therefore broaden audiological service delivery to more patients. Furthermore, it could utilize vast patient databases, which presently lie dormant in most clinical settings.…”