A Cascade Spatial-Temporal Denoising Strategy for Cardiac Artifact Removal in Spinal Cord Electrophysiology: A Pilot Study

Background: Cardiac electrical activity introduces severe artifacts in spinal cord electrophysiology (ESG), masking the subtle neural signals. Traditional methods, such as Spatial Signal Projection (SSP) and Adaptive Template Subtraction (ATS), are often used individually but have inherent limitations. Objective: This study aims to propose and validate a novel cascade strategy (SSP followed by ATS) to optimize cardiac artifact removal. Methods: Raw ESG data from three subjects were obtained from the OpenNeuro public dataset (ds004388), resulting in a total of 117 channels (116 spinal channels plus 1 left mastoid reference channel, LM). The proposed cascade strategy was compared against standalone SSP and ATS. The Residual Index (RI) metric was used to quantify denoising performance. Wilcoxon signed-rank tests showed significant improvement over both SSP and ATS (both p < 0.001); LMM accounting for subject random effects was consistent with the superiority over ATS (p < 0.001) and SSP (p < 0.05). Results: The cascade strategy achieved the lowest RI across 116 out of 117 channels (99.1%). Furthermore, qualitative waveform visualization suggested that the cascade method suppressed cardiac spikes while largely preserving underlying low-frequency neural signals. Conclusion: The cascade spatial-temporal strategy represents a promising and effective approach in this pilot dataset for ESG denoising, providing a potentially useful baseline for future brain-spine interface (BSI) decoding.