Unsupervised and real-time spike sorting chip for neural signal processing in hippocampal prosthesis

Area of Science:

• Neuroscience
• Biomedical Engineering
• Electrical Engineering

Background:

• Hippocampal damage causes memory loss and cognitive disorders with no current effective medical treatment.
• Hippocampal prostheses are being developed to restore function by replacing damaged tissue.
• Neural signal processing, specifically spike sorting, is crucial for hippocampal prosthesis operation.

Purpose of the Study:

• To develop a 16-channel spike sorting chip for hippocampal prostheses.
• To improve the detection and classification of neural spikes for prosthetic applications.
• To address the limitations of existing spike sorting methods in terms of accuracy and handling complex neural data.

Main Methods:

• A 16-channel spike sorting processor was designed with independent channels.
• An automatic threshold estimation method for the Osort clustering algorithm was developed for hardware implementation.
• A novel distance metric and an optimized Bayes optimal template matching algorithm with preselection were introduced.

Main Results:

• The chip was fabricated using 40-nm CMOS technology, achieving a core area of 0.0175 mm²/ch and power consumption of 19.0 μW/ch.
• Evaluation using synthetic and realistic data demonstrated high performance.
• The proposed chip achieved superior detection and classification accuracy compared to existing spike sorting processors.

Conclusions:

• A 16-channel spike sorting chip for hippocampal prostheses has been successfully developed.
• The chip enables unsupervised clustering and real-time detection and classification of neural spikes.
• The developed chip can effectively handle partially overlapping spikes, a capability not reported in previous work.