A fast and simple algorithm for accurate spike detection in HD-MEA recordings.

Area of Science:

• Neuroscience
• Computational Biology
• Bioengineering

Background:

• High-density microelectrode arrays (HD-MEAs) are crucial for studying neuronal activity and network dynamics.
• Analyzing large, complex datasets from HD-MEAs presents significant challenges.
• Existing spike detection methods, like Root-Mean-Square (RMS), underestimate spikes during neuronal bursting and spike sorting is computationally intensive.

Purpose of the Study:

• To develop and validate a more accurate and computationally efficient spike detection and de-duplication method for HD-MEA data.
• To address the limitations of RMS-based methods in capturing burst firing and the computational burden of spike sorting.

Main Methods:

• Optimized a scaled median of absolute deviations (MED) based detection method for improved accuracy during high firing rates.
• Integrated a de-duplication (DP) step to handle spikes detected across multiple electrodes, enhancing MED's accuracy.
• Developed the combined de-duplication and MED (DP-MED) method, comparing its performance and computational cost against Kilosort-4.

Main Results:

• The MED-based method detected over 50% more spikes during burst periods compared to the RMS-based method.
• Simulated data showed DP-MED offered higher precision than Kilosort-4, with slightly lower accuracy at increasing firing rates.
• In cortical cultures, DP-MED detected a similar number of spikes as Kilosort-4 but was 40 times faster.

Conclusions:

• DP-MED effectively detects spikes missed by RMS methods, particularly during neuronal bursting.
• The DP-MED method provides a computationally efficient alternative to Kilosort-4 for analyzing HD-MEA data.
• DP-MED demonstrates significant utility for applications like drug screening using HD-MEAs due to its speed and accuracy.