Optimizing the Yield of Multi-Unit Activity by Including the Entire Spiking Activity

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Neurophysiological recordings often suffer from low signal-to-noise ratio (SNR), hindering accurate detection of neuronal responses.
• Conventional spike detection methods relying on thresholds fail in poor SNR conditions, limiting data analysis.
• Automated analysis is crucial for real-time applications like brain-computer interfaces.

Purpose of the Study:

• To introduce and evaluate a novel threshold-independent method for analyzing low SNR neurophysiological data.
• To assess the sensitivity and reliability of the Entire Spiking Activity (ESA) signal for detecting evoked responses.
• To compare ESA-based detection with conventional Multi-Unit Activity (MUA) detection in macaque V1 recordings.

Main Methods:

• Developed a fast, automated procedure involving full-wave rectification and low-pass filtering to derive the ESA signal.
• Applied an automated receptive field (RF) mapping procedure to semi-chronic recordings from macaque primary visual cortex (V1).
• Compared RF detection rates, selectivity, size, and orientation tuning between ESA and MUA-based methods across different SNR levels.

Main Results:

• ESA improved RF detection rates by 2.5x in low SNR data and yielded 30% more RFs in medium/high SNR data compared to MUA.
• ESA-based RF detection remained superior even when using optimized individual spike thresholds for MUA.
• ESA showed comparable selectivity measures, larger RFs, and more orientation-tuned sites than MUA, with similar preferred orientations.

Conclusions:

• The ESA signal is a powerful tool for automated, fast, and reliable neural response detection, especially in low SNR conditions.
• ESA's high sensitivity and independence from user intervention make it suitable for brain-computer interfaces and neuroprosthetics.
• ESA preserves full spiking activity information, offering a valuable alternative for offline analysis of limited SNR data.