SpikeSift: a computationally efficient and drift-resilient spike sorting algorithm.

Area of Science:

• Computational neuroscience
• Electrophysiology
• Signal processing

Background:

• Spike sorting is crucial for analyzing extracellular recordings to isolate single-neuron activity.
• Existing methods struggle with overlapping spikes and recording instabilities like electrode drift.
• Current algorithms often fail to balance accuracy with the computational efficiency needed for large datasets.

Purpose of the Study:

• Introduce SpikeSift, a novel spike-sorting algorithm designed for high accuracy and computational efficiency.
• Address the challenges of electrode drift and overlapping spikes in extracellular recordings.
• Provide a practical tool for analyzing large-scale neural datasets on standard hardware.

Main Methods:

• SpikeSift partitions recordings into stationary segments to mitigate drift.
• It employs an iterative detect-and-subtract scheme for simultaneous spike detection and clustering.
• A template-alignment stage preserves neuronal identity across segments without continuous trajectory estimation.

Main Results:

• SpikeSift matches or surpasses the accuracy of state-of-the-art spike-sorting methods.
• The algorithm is an order of magnitude faster than existing methods on a single CPU core.
• Validation was performed on intracellularly validated datasets and biophysically realistic simulations.

Conclusions:

• SpikeSift offers a robust solution for accurate and efficient spike sorting, even with challenging data.
• Its drift resilience and computational efficiency make it broadly accessible for neurophysiological research.
• The algorithm preserves data quality for downstream analysis, enhancing the utility of extracellular recordings.