Multi-unit activity (MUA) reflects the combined firing of local neuronal populations and is widely used to study cortical dynamics. However, MUA estimation is often distorted by artifacts introduced during electrical micro-stimulation, including abrupt signal deflections and filter-induced ringing, which can lead to false spike detections and misinterpretations of neural responses. In this study, we introduce a method to minimize such artifacts by applying a local cubic polynomial to interpolate over stimulation-contaminated segments, preserving the continuity of the signal while preventing oscillations. Using intra-cortical recordings from a Utah Electrode Array (UEA), we show that polynomial correction effectively reduces artifact-driven peaks and maintains physiological activity patterns. Comparison between corrected and uncorrected signals shows a smoother, more plausible MUA profile, with a marked reduction in pre- and post-stimulus activity. Our approach enables reliable, high-resolution assessment of neuronal population responses during stimulation, providing a practical framework for peri-stimulus analyses and supporting robust interpretation of cortical activation in intra-cortical micro-stimulation experiments.