neuropixels-analysis

Spike sorting on Neuropixels data requires loading SpikeGLX or Open Ephys recordings via SpikeInterface, preprocessing with filtering and common referencing, running sorters like Kilosort4, and computing quality metrics to classify curated units.

Motion correction in extracellular ephys recordings involves detecting peaks in SpikeInterface, estimating probe drift across channels, and applying motion correction algorithms before executing spike sorting to ensure accurate unit separation.

Yes, SpikeInterface supports threshold-based, model-based, or AI-assisted curation of sorted units and exports curated results to Phy for manual inspection, NWB format for sharing, or generated tables and visualization reports.

NWB workflows are fully supported for preprocessing raw Neuropixels data, including filtering, bad-channel handling, common referencing, drift correction, spike sorting execution, and final export of curated unit classifications back to NWB.

Quality metrics for unit assessment are computed in SpikeInterface after spike sorting by evaluating waveform shapes, firing rates, and contamination scores to distinguish well-isolated single units from multi-unit or noise clusters.

CPU-based spike sorting alternatives process Neuropixels recordings without GPU acceleration, potentially limiting throughput on high-density probes compared to Kilosort4, though SpikeInterface provides compatible options for various hardware environments.