neuropixels-analysis

Neuropixels data analysis processes raw recordings through preprocessing, motion correction, spike sorting with Kilosort4, SpykingCircus2, or Mountainsort5, then applies quality metrics and curation. Load data via SpikeInterface (SpikeGLX, Open Ephys, NWB formats), run the end-to-end workflow, and export curated units to Phy or NWB for publication.

The workflow supports Kilosort4, Kilosort3, SpykingCircus2, and Mountainsort5, with ensemble comparison capabilities. SpikeInterface abstracts these sorters, allowing you to run multiple algorithms on the same preprocessed data and evaluate trade-offs before finalizing curated units.

Yes, the workflow handles both Neuropixels 1.0 and 2.0 data and supports multiple input formats: SpikeGLX, Open Ephys, and NWB. Probeinterface integration enables probe-specific configuration and geometry handling for accurate preprocessing and sorting.

Drift detection and correction occur during preprocessing using phase shift and motion correction techniques. The workflow identifies and compensates for electrode drift across the recording session, improving spike sorting accuracy and unit quality before sorting algorithms are applied.

The postprocessing step computes comprehensive quality metrics including spike amplitudes, waveform properties, templates, and Allen/IBL-style curation criteria. These metrics support both automated filtering and AI-assisted visual analysis to identify publication-ready units.

The workflow exports to Phy and NWB natively. NWB is a standard format compatible with downstream neuroscience analysis tools; Phy enables interactive curation and refinement before final export or archival to other formats via standard conversion pipelines.