Evaluation of Spike Sorting Algorithms: Application to Human Subthalamic Nucleus Recordings and Simulations

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• Accurate single-unit activity (SUA) extraction from multi-unit signals is essential for analyzing neural synchrony in extracellular recordings.
• Spike sorting algorithms are used to differentiate neuronal origins but yield inconsistent results, impacting downstream analyses.
• Subthalamic nucleus recordings from Parkinson's disease patients present unique challenges for spike sorting due to signal complexity.

Purpose of the Study:

• To identify the optimal spike sorting algorithm for subthalamic nucleus recordings in Parkinson's disease patients.
• To evaluate the performance of prevalent spike sorting algorithms using both experimental and artificial data.
• To assess the influence of different algorithms on single-unit assignments and firing characteristic estimations.

Main Methods:

• Application and evaluation of multiple spike sorting algorithms from 'Plexon Offline Sorter' on experimental data (ED).
• Validation using artificial data (AD) with known ground truth, including single units with varying shape similarity and background noise.
• Comparative analysis of sorting consistency, variability, and impact on firing characteristics across different algorithms.

Main Results:

• Spike sorting algorithm choice significantly influences single-unit assignments and subsequent analyses.
• High variability in sorting results was observed, increasing with the similarity in shape between single units.
• Significant differences in estimated firing characteristics were found depending on the algorithm used.

Conclusions:

• Valley-Seeking algorithms provide superior accuracy when artifact exclusion is critical.
• K-Means algorithm is a preferable option for 'clean' data where artifact exclusion is less prioritized.
• Standardized validation procedures using ground truth data are necessary to ensure reliable spike sorting and analysis.