A compact representation for the auditory full-range response and its fast denoising using an image filter based on the Radon Transform

Abstract

The Auditory Brainstem, Middle-Latency and Late Responses, a class of event-related potentials (ERPs), are of considerable interest in neuroscience research as robust neural correlates of different processing stages along the auditory pathway. While most research to date centers around one of the responses at a time for practical reasons, recent efforts indicate a paradigm shift towards acquiring them together, enabling the simultaneous monitoring of all auditory processing stages from the brainstem to the cortex. In this paper, we introduce a compact representation for this Auditory Full-Range Response (AFRR) as an ERP map with adaptive sampling rate, making it suitable for computationally inexpencive image filtering. Furthermore, we propose a novel algorithm for the fast denoising of such ERP maps based on the Radon Transform and its inversion by filtered backprojection. Its performance is compared qualitatively to a Gaussian means filter using a real-world chirp-evoked AFRR recording. The algorithm exhibits good noise suppression as well as high preservance of the single-response structure, making it a promising denoising tool for future ERP studies.