Classification of overlapping spikes using convolutional neural networks and long short term memory

Area of Science:

• Neuroscience
• Computational Neuroscience
• Machine Learning

Background:

• Spike sorting is crucial for analyzing neural activity.
• Accurate separation of overlapping neuronal spikes remains a challenge.
• Existing methods may struggle with complex spike data.

Purpose of the Study:

• To develop and evaluate a novel deep learning approach for overlapping spike sorting.
• To improve the accuracy and efficiency of neuronal spike separation.
• To address limitations of previous deep learning methods in spike sorting.

Main Methods:

• Utilized a hybrid deep learning model combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM).
• Trained and validated the model using simulated electrophysiological data.
• Evaluated performance on experimental data from a macaque monkey's primary visual cortex.

Main Results:

• Achieved high clustering accuracy: >99.9% for non-overlapping spikes and >99.0% for overlapping spikes on simulated data.
• Demonstrated superior performance compared to a previous 1D-CNN deep learning approach.
• Successfully isolated overlapping spikes from two to five simultaneously recorded neurons in experimental data.

Conclusions:

• The proposed CNN + LSTM method offers significant advantages for high-accuracy overlapping spike sorting.
• This approach provides a robust foundation for tackling more complex neurophysiological analysis tasks.
• The method shows promise for distinguishing simultaneous recordings from multichannel neuronal activities.