SpikeSift: A Computationally Efficient and Drift-Resilient Spike Sorting Algorithm

Spike sorting is a fundamental step in analyzing extracellular recordings, enabling the isolation of individual neuronal activity, yet it remains a challenging problem due to overlapping signals and recording instabilities, including electrode drift. While numerous algorithms have been developed to address these challenges, many struggle to balance accuracy and computational efficiency, limiting their applicability to largescale datasets. In response, we introduce SpikeSift, a novel spike sorting algorithm designed to mitigate drift by partitioning recordings into short, relatively stationary segments, with spikes subsequently sorted within each. To preserve neuronal identity across segment boundaries, a computationally efficient alignment process merges clusters without relying on continuous trajectory estimation. In contrast to conventional methods that separate spike detection from clustering, SpikeSift integrates these processes within an iterative detect-andsubtract framework, enhancing clustering accuracy while maintaining computational efficiency. Evaluations on intracellularly validated datasets and biophysically realistic MEArec simulations confirm that SpikeSift maintains high sorting accuracy even in the presence of electrode drift, providing a scalable and computationally efficient solution for large-scale extracellular recordings