The auditory brainstem response (ABR) is a voltage response evoked by acoustic stimuli as sound is processed along the auditory pathway. It consists of electrical signals resulting from the sum of sound-evoked activity along the auditory nerve and brainstem nuclei. ABR analysis determines the sound intensity at which a neural response first appears (hearing threshold) . Previous studies in rats in mice have shown that ABR thresholds do not indicate absolute behavioural hearing thresholds [2, 3]. However, ABR audiometry has been used extensively in animal hearing research for examining gene therapy [4–10], cell-replacement therapies [11–14], and noise-induced hearing loss [15–20].
The ABR offers an objective measurement of auditory signal processing. The objectivity is diminished by conventional visual inspection of the ABR threshold level. Subjectivity and variability are introduced when the investigator has to decide when a complex, multi-component response first becomes distinguishable from background noise . Methodologies have been developed to address the subjective component of threshold detection by including criteria about the shape, pattern, or absolute amplitude of the response, yet these still require a visual decision about the presence of a signal. Eliminating subjectivity in auditory threshold determination would improve the sensitivity and reliability of this important audiometric technique.
While visual estimation remains the conventional technique for ABR threshold detection, a need for automated statistical methods for detection is highly recognised. Several methods have been developed based on the techniques of Fsp analysis [22–25], cross correlation [26–28] and feature vectors [29–32]. Fsp analysis requires calculation of a variance ratio in the ABR waveform followed by application of the F-statistic to this ratio. Cross correlation measures the degree of similarity between a sliding template and an averaged waveform. Feature vectors quantify selected components of the response's time course. Fsp has been incorporated into available software (Compumedics Ltd) yet the other methods lack comparable implementation.
Here we develop a simple, fully automated auditory threshold detection method to address the subjectivity and variability associated with visual estimation of ABRs. This method is based on the signal-to-noise ratio and the software has been made readily available . The algorithm is calibrated by comparison with visual estimation, implemented via investigation of stem cell transplantation, and compared against variability obtained with visual estimation.