A Denoising Algorithm Based on Spatial and Spectral Domain Filtering for Electrically Evoked Compound Action Potential Enhancement

The electrically evoked compound action potential (ECAP) is a physiological signal that is important for clinicians to assess a patient's hearing. A clean ECAP can be used to estimate the patient’s auditory neural activity patterns and ECAP magnitude using the panoramic ECAP (PECAP) approach. However, noisy ECAPs can introduce significant errors in parameter estimation, especially at low signal-to-noise ratios (SNRs). To address this problem, a two-stage preprocessing denoising (TSPD) algorithm is developed. First, the ECAP matrix can be constructed using the forward-masking technique. Thus, the ECAP matrix can be viewed as an image for the first stage of noise reduction using the improved spatial median filter. Then, the denoised ECAP matrix can be expanded and rearranged into a vector for the second stage of noise reduction using the log-spectral amplitude (LSA) Wiener filter. Lastly, the enhanced ECAP vector is reconstructed to the ECAP matrix for aforementioned parameter estimation using PECAP. The results show that the proposed method, called PECAP-TSPD, improves the estimation of auditory neural activity patterns and ECAP magnitudes compared to those of the baseline (PECAP) and unprocessed ECAP signals in terms of the normalized root mean square error (RMSE).