A Novel Algorithm to Derive Spread of Excitation Based on Deconvolution

Objective: The width of the spread of excitation (SOE) curve has been widely thought to represent an estimate of SOE. Therefore, correlates between psychophysical parameters, such as pitch discrimination and speech perception, and the width of SOE curves, have long been investigated. However, to date, no relationships between these objective and subjective measurements have yet been determined. In a departure from the current thinking, the authors now propose that the SOE curve, recorded with forward masking, is the equivalent of a convolution operation. As such, deconvolution would be expected to retrieve the excitation areas attributable to either masker or probe, potentially more closely revealing the actual neural SOE. This study aimed to develop a new analytical tool with which to derive SOE using this principle. Design: Intraoperative SOE curve measurements of 16 subjects, implanted with an Advanced Bionics implant, were analyzed. Evoked compound action potential (ECAP)-based SOE curves were recorded on electrodes 3 to 16, using the forward masker paradigm, with variable masker. The measured SOE curves were then compared with predicted SOE curves, built by the convolution of basic excitation density profiles (EDPs). Predicted SOE curves were fitted to the measured SOEs by iterative adjustment of the EDPs for the masker and the probe. Results: It was possible to generate a good fit between the predicted and measured SOE curves, inclusive of their asymmetry. The rectangular EDP was of least value in terms of its ability to generate a good fit; smoother SOE curves were modeled using the exponential or Gaussian EDPs. In most subjects, the EDP width (i.e., the size of the excitation area) gradually changed from wide at the apex of the electrode array, to narrow at the base. A comparison of EDP widths to SOE curve widths, as calculated in the literature, revealed that the EDPs now provide a measure of the SOE that is qualitatively distinct from that provided using conventional methods. Conclusions: This study shows that an eCAP-based SOE curve, measured with forward masking, can be treated as a convolution of EDPs for masker and probe. The poor fit achieved for the measured and modeled data using the rectangular EDP, emphasizes the requirement for a sloping excitation area to mimic actual SOE recordings. Our deconvolution method provides an explanation for the frequently observed asymmetry of SOE curves measured along the electrode array, as this is a consequence of a wider excitation area in the apical part of the cochlea, in the absence of any asymmetry in the actual EDP. In addition, broader apical EDPs underlie the higher eCAP amplitudes found for apical stimulation.