An automatic spike sorting algorithm based on adaptive spike detection and a mixture of skew-t distributions.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• High-density electrodes enable recording neuronal ensembles, crucial for understanding neural circuits.
• Chronic implantation leads to brain tissue changes, causing spike waveform distortion and unstable spike sorting.
• Existing automatic spike sorting algorithms struggle with unstable spike shapes, leading to spike loss and inaccurate analysis.

Purpose of the Study:

• To develop a robust automatic spike sorting algorithm addressing waveform distortion and instability.
• To improve spike detection and clustering for accurate neural data analysis.
• To provide a generalizable solution for precise spike sorting in various experimental conditions.

Main Methods:

• Developed an automatic spike sorting algorithm utilizing adaptive spike detection and a mixture of skew-t distributions.
• Implemented an adaptive detection procedure with multi-point alignment and statistical filtering for accurate spike identification.
• Employed a mixture of skew-t distributions for robust clustering of distorted and non-symmetrical spike waveforms.

Main Results:

• The proposed algorithm significantly enhances spike sorting precision and recall across various signal-to-noise ratios.
• Adaptive detection effectively filters out mistakenly detected spikes, reducing spike loss.
• Clustering using skew-t distributions successfully handles non-symmetrical clusters and improves overall sorting accuracy.

Conclusions:

• The developed algorithm offers a robust solution for precise spike sorting despite neural signal distortions.
• The method demonstrates generalizability and effectiveness on diverse in vitro and in vivo datasets.
• This work advances the analysis of large-scale neural recordings for deeper insights into neuronal circuit mechanisms.