Synthetic data-driven overlapped neural spikes sorting: decomposing hidden spikes from overlapping spikes.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Spike sorting is crucial for analyzing neural activity and understanding neural coding.
• Overlapping spikes, caused by simultaneous neuronal firing, pose a significant challenge in accurate spike sorting.
• Existing methods struggle to identify individual neurons due to overlapping signals and lack of ground truth.

Purpose of the Study:

• To develop a robust method for identifying and decomposing overlapping spikes in extracellular neuronal recordings.
• To improve the accuracy of neural activity analysis by recovering hidden spike information.
• To enable precise inference of neuronal synchronization through spike decomposition.

Main Methods:

• Spike waveform templates are estimated using discriminative subspace learning and the isolation forest algorithm.
• Synthetic spikes are generated from estimated templates to train a classifier for identifying overlapping spikes.
• Particle swarm optimization is employed to decompose identified overlapping spikes into individual hidden spikes.

Main Results:

• The proposed method achieved a maximum F1 score of 0.88 in accurately identifying overlapping spikes using synthetic data.
• The approach successfully decomposes overlapping spikes, allowing for the recovery of hidden neuronal signals.
• Inferred synchronization patterns between hidden spikes were achieved by reallocating decomposed spikes into distinct clusters.

Conclusions:

• This study presents an effective computational approach for addressing the challenge of overlapping spikes in neural recordings.
• The developed method enhances the precision of spike sorting, offering a valuable tool for neuroscientific research.
• Accurate identification and decomposition of overlapping spikes facilitate a deeper understanding of neural coding and synchronization.