Neuropixels AnalysisCAUTION

The largest open-source medical AI skills library for OpenClaw🦞.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: neuropixels-analysis
description: "Neuropixels neural recording analysis. Load SpikeGLX/OpenEphys data, preprocess, motion correction, Kilosort4 spike sorting, quality metrics, Allen/IBL curation, AI-assisted visual analysis, for Neuropixels 1.0/2.0 extracellular electrophysiology. Use when working with neural recordings, spike sorting, extracellular electrophysiology, or when the user mentions Neuropixels, SpikeGLX, Open Ephys, Kilosort, quality metrics, or unit curation."
---

# Neuropixels Data Analysis

## Overview

Comprehensive toolkit for analyzing Neuropixels high-density neural recordings using current best practices from SpikeInterface, Allen Institute, and International Brain Laboratory (IBL). Supports the full workflow from raw data to publication-ready curated units.

## When to Use This Skill

This skill should be used when:
- Working with Neuropixels recordings (.ap.bin, .lf.bin, .meta files)
- Loading data from SpikeGLX, Open Ephys, or NWB formats
- Preprocessing neural recordings (filtering, CAR, bad channel detection)
- Detecting and correcting motion/drift in recordings
- Running spike sorting (Kilosort4, SpykingCircus2, Mountainsort5)
- Computing quality metrics (SNR, ISI violations, presence ratio)
- Curating units using Allen/IBL criteria
- Creating visualizations of neural data
- Exporting results to Phy or NWB

## Supported Hardware & Formats

| Probe | Electrodes | Channels | Notes |
|-------|-----------|----------|-------|
| Neuropixels 1.0 | 960 | 384 | Requires phase_shift correction |
| Neuropixels 2.0 (single) | 1280 | 384 | Denser geometry |
| Neuropixels 2.0 (4-shank) | 5120 | 384 | Multi-region recording |

| Format | Extension | Reader |
|--------|-----------|--------|
| SpikeGLX | `.ap.bin`, `.lf.bin`, `.meta` | `si.read_spikeglx()` |
| Open Ephys | `.continuous`, `.oebin` | `si.read_openephys()` |
| NWB | `.nwb` | `si.read_nwb()` |

## Quick Start

### Basic Import and Setup

```python
import spikeinterface.full as si
import neuropixels_analysis as npa

# Configure parallel processing
job_kwargs = dict(n_jobs=-1, chunk_duration='1s', progress_bar=True)
```

### Loading Data

```python
# SpikeGLX (most common)
recording = si.read_spikeglx('/path/to/data', stream_id='imec0.ap')

# Open Ephys (common for many labs)
recording = si.read_openephys('/path/to/Record_Node_101/')

# Check available streams
streams, ids = si.get_neo_streams('spikeglx', '/path/to/data')
print(streams)  # ['imec0.ap', 'imec0.lf', 'nidq']

# For testing with subset of data
recording = recording.frame_slice(0, int(60 * recording.get_sampling_frequency()))
```

### Complete Pipeline (One Command)

```python
# Run full analysis pipeline
results = npa.run_pipeline(
    recording,
    output_dir='output/',
    sorter='kilosort4',
    curation_method='allen',
)

# Access results
sorting = results['sorting']
metrics = results['metrics']
labels = results['labels']
```

## Standard Analysis Workflow

### 1. Preprocessing

```python
# Recommended preprocessing chain
rec = si.highpass_filter(recording, freq_min=400)
rec = si.phase_shift(rec)  # Required for Neuropixels 1.0
bad_ids, _ = si.detect_bad_channels(rec)
rec = rec.remove_channels(bad_ids)
rec = si.common_reference(rec, operator='median')

# Or use our wrapper
rec = npa.preprocess(recording)
```

### 2. Check and Correct Drift

```python
# Check for drift (always do this!)
motion_info = npa.estimate_motion(rec, preset='kilosort_like')
npa.plot_drift(rec, motion_info, output='drift_map.png')

# Apply correction if needed
if motion_info['motion'].max() > 10:  # microns
    rec = npa.correct_motion(rec, preset='nonrigid_accurate')
```

### 3. Spike Sorting

```python
# Kilosort4 (recommended, requires GPU)
sorting = si.run_sorter('kilosort4', rec, folder='ks4_output')

# CPU alternatives
sorting = si.run_sorter('tridesclous2', rec, folder='tdc2_output')
sorting = si.run_sorter('spykingcircus2', rec, folder='sc2_output')
sorting = si.run_sorter('mountainsort5', rec, folder='ms5_output')

# Check available sorters
print(si.installed_sorters())
```

### 4. Postprocessing

```python
# Create analyzer and compute all extensions
analyzer = si.create_sorting_analyzer(sorting, rec, sparse=True)

analyzer.compute('random_spikes', max_spikes_per_unit=500)
analyzer.compute('waveforms', ms_before=1.0, ms_after=2.0)
analyzer.compute('templates', operators=['average', 'std'])
analyzer.compute('spike_amplitudes')
analyzer.compute('correlograms', window_ms=50.0, bin_ms=1.0)
analyzer.compute('unit_locations', method='monopolar_triangulation')
analyzer.compute('quality_metrics')

metrics = analyzer.get_extension('quality_metrics').get_data()
```

### 5. Curation

```python
# Allen Institute criteria (conservative)
good_units = metrics.query("""
    presence_ratio > 0.9 and
    isi_violations_ratio < 0.5 and
    amplitude_cutoff < 0.1
""").index.tolist()

# Or use automated curation
labels = npa.curate(metrics, method='allen')  # 'allen', 'ibl', 'strict'
```

### 6. AI-Assisted Curation (For Uncertain Units)

When using this skill with Claude Code, Claude can directly analyze waveform plots and provide expert curation decisions. For programmatic API access:

```python
from anthropic import Anthropic

# Setup API client
client = Anthropic()

# Analyze uncertain units visually
uncertain = metrics.query('snr > 3 and snr < 8').index.tolist()

for unit_id in uncertain:
    result = npa.analyze_unit_visually(analyzer, unit_id, api_client=client)
    print(f"Unit {unit_id}: {result['classification']}")
    print(f"  Reasoning: {result['reasoning'][:100]}...")
```

**Claude Code Integration**: When running within Claude Code, ask Claude to examine waveform/correlogram plots directly - no API setup required.

### 7. Generate Analysis Report

```python
# Generate comprehensive HTML report with visualizations
report_dir = npa.generate_analysis_report(results, 'output/')
# Opens report.html with summary stats, figures, and unit table

# Print formatted summar

CAUTIONgrade B · trust 89/100 Install with care. The audit found things worth knowing before you trust its output.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (3)

ax.set_ylabel('Depth (μm)')

print(f"  Estimated drift range: {drift_estimate:.1f} μm")

print(f"Drift > 20 μm detected, applying motion correction...")

Gates applied: no_behavioural_pass.