A Clinical Feasibility Study of Spinal Evoked Compound Action Potential Estimation Methods

Abstract

Objectives: Spinal cord stimulation (SCS) is a treatment for chronic neuropathic pain. Recently, SCS has been enhanced further with evoked compound action potential (ECAP) sensing. Characteristics of the ECAP, if appropriately isolated from concurrent stimulation artifact (SA), may be used to control, and aid in the programming of, SCS systems. Here, we characterize the sensitivity of the ECAP growth curve slope (S) to both neural response (|Sresp|) and SA contamination (|Sart|) for four spinal ECAP estimation methods with a novel performance measure (|Sresp/Sart|). Materials and Methods: We collected a library of 112 ECAP and associated artifact recordings with swept stimulation amplitudes from 14 human subjects. We processed the signals to reduce SA from these recordings by applying one of three schemes: a simple high-pass (HP) filter, subtracting an artifact model (AM) consisting of decaying exponential and linear components, or applying a template correlation method consisting of a triangularly weighted sinusoid. We compared these against each other and to P2-N1, a standard method of measuring ECAP amplitude. We then fit the ECAP estimates from each method with a function representing the growth curve and calculated the Sresp and Sart parameters following the fit. Results: Any SA reduction scheme selected may result in under- or overestimation of neural activation or misclassification of SA as ECAP. In these experiments, the ratio of neural signal preservation to SA misclassification (|Sresp/Sart|) on the ECAP estimate was superior (p < 0.05) with the HP and AM schemes relative to the others. Conclusions: This work represents the first comprehensive assessment of spinal ECAP estimation schemes. Understanding the clinically relevant sensitivities of these schemes is increasingly important, particularly with closed-loop SCS systems using ECAP as a feedback control variable where misclassification of artifact as neural signal may lead to suboptimal therapy adjustments.