Spike Sorting

Spike sorting is the process of identifying which neuron produced each action potential in your recording. Since a single electrode detects spikes from multiple nearby neurons, sorting separates these into single-unit activity for individual cells.

Why it matters: Spike sorting transforms multi-unit recordings into single-neuron data, which is essential for understanding how individual neurons encode information, respond to stimuli, and interact with other neurons.

How it works:

1. Detection: Find times when voltage crosses threshold (potential spikes)
2. Feature extraction: Measure spike waveform characteristics (amplitude, width, PCA components)
3. Clustering: Group similar waveforms together (each cluster = one neuron)
4. Validation: Check cluster quality, remove noise, verify single-unit criteria

Automated sorters: Modern tools (Kilosort, MountainSort, SpyKING CIRCUS) use sophisticated algorithms to sort thousands of spikes across hundreds of channels. Manual curation is still recommended for critical analyses.

Quality metrics:

• Isolation distance: How separable is this cluster from others?
• SNR: Signal-to-noise ratio of spike amplitude
• Refractory violations: Biological neurons can't fire faster than ~1 ms

Challenges: Overlapping spikes, electrode drift, low SNR, and similar waveforms from different neurons make sorting difficult. High-density probes help by providing multiple spatial views of each neuron.