Synthetic Simulations of Extracellular Recordings for Spike Sorting Algorithm Evaluation

10 minutes of simulated extracellular recordings contain the activity of one multi-unit and two single-units across different firing rates and signal-to-noise ratios. The dataset includes 14,400,000 sampled voltage values at 24 kHz and labels for spiking events, constructed from 594 average spike shapes derived from real monkey neocortex and basal ganglia recordings. Timothy Zhang from the University of Toronto amalgamated these simulation files, which were originally created by Martinez et al. in 2009.

Use Cases

• Benchmarking spike sorting algorithm performance based on labeled spiking events with varying SNR and firing rates.
• Training machine learning models for spike detection and classification based on synthetic voltage traces.
• Studying the effect of signal-to-noise ratio on unit isolation accuracy based on the provided amplitude (SNR) labels.
• Simulating realistic extracellular recording scenarios for methodological development based on shapes from real monkey recordings.

Strengths

• Contains 14,400,000 sampled voltage values from a 10-minute simulation at a 24 kHz sampling rate.
• Spike shapes are derived from a database of 594 different average shapes taken from real monkey recordings.
• Labels include timestep, spike class, amplitude (SNR), and firing rate for each event, supporting detailed evaluation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• The dataset is synthetic, which may not capture all complexities of in vivo recordings.

Provenance

Source

University of Toronto (Timothy Zhang), amalgamating simulations originally from the Neuroengineering Lab at the University of Leicester.

Collection Method

Synthetic simulations combining spike shapes from real recordings.