Unsupervised spike sorting for multielectrode arrays based on spike shape features and location methods.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Bioengineering

Background:

• Microelectrode arrays (MEAs) facilitate simultaneous recording of neuronal activity, crucial for understanding complex neural networks.
• Spike sorting, essential for analyzing extracellular recordings, faces accuracy challenges with dense spike data from MEAs.

Purpose of the Study:

• To introduce an unsupervised spike sorting pipeline, UMAP-COM, that improves accuracy for MEA data.
• To combine Uniform Manifold Approximation and Projection (UMAP) for spike shape features and Center of Mass (COM) for spike localization.

Main Methods:

• Developed an unsupervised pipeline integrating UMAP for dominant spike shape feature extraction.
• Incorporated COM for precise spike location estimation.
• Validated the UMAP-COM method on diverse, publicly available MEA datasets.

Main Results:

• The UMAP-COM method demonstrated superior accuracy compared to existing spike sorting techniques.
• UMAP was identified as an effective method for extracting representative spike shape features.
• COM proved to be a more accurate spike localization method, enhancing overall sorting performance.

Conclusions:

• The UMAP-COM pipeline offers a significant advancement in unsupervised spike sorting for MEA recordings.
• The combination of UMAP and COM effectively addresses the limitations of current methods, improving neuronal activity analysis.
• This approach provides a robust tool for neuroscientists studying neuronal network dynamics.