Physiological Parameter Estimation for Dorsal Column Spinal Cord Stimulation

Spinal Cord Stimulation (SCS) is an established treatment option for patients living with chronic neuropathic pain. Although recently developed implanted pulse generator (IPG) systems are utilizing real-time electrophysiological measurements of evoked compound action potential (ECAP) amplitude as a feedback signal for modulating SCS therapy, it is not clear whether this is the optimal feedback signal for maintaining consistent neural activation. In this paper, we consider ECAP amplitude alongside other extracted features, such as AUC, N1 time, and conduction velocity, and investigate how these features respond to changes in physiological parameters in a computational model of ECAPs produced by dorsal column SCS. We use a simulated test bed to compare therapy-relevant parameter estimation by linear estimators and a Kalman Filter constructed from different ECAP features, and we demonstrate that a Kalman Filter using N1 time or conduction velocity has robust performance across the range of simulated conditions at the stimulating and sensing electrodes, while estimators using ECAP amplitude and AUC features are shown to be prone to higher error when conditions at the stimulating and sensing electrodes are not ideal. These results may drive future adaptive SCS therapy developments and a reconsideration of how to leverage extracted ECAP features for detecting therapy-relevant signal changes.

I. Introduction

Dorsal column spinal cord stimulation (SCS) has been an established therapy for chronic neuropathic pain for decades [1]. Recent developments in the field have seen an expansion of therapy modalities provided to patients, including the use of various stimulation patterns [2] and approaches to sub-perception therapy delivery [3]. Additionally, advancements in implanted pulse generator (IPG) technology have afforded therapy adaptation approaches based on on-board accelerometers [4] and electrophysiological measurements [5] in efforts to maintain consistent neural activation.