技能 neuropixels-analysis
🧠

neuropixels-analysis

安全 ⚙️ 外部命令📁 檔案系統存取🌐 網路存取

分析 Neuropixels 神經記錄

也可從以下取得: davila7

此技能提供對 Neuropixels 高密度神經記錄的全面分析。它處理從原始數據加載到使用 SpikeInterface 和 Kilosort4 算法生成可直接發表的精選單元的完整工作流程。

支援: Claude Codex Code(CC)
🥈 81 白銀
1

下載技能 ZIP

2

在 Claude 中上傳

前往 設定 → 功能 → 技能 → 上傳技能

3

開啟並開始使用

測試它

正在使用「neuropixels-analysis」。 Load my Neuropixels recording and run the complete analysis pipeline

預期結果:

  • Recording: 384 channels, 600.2 seconds
  • Preprocessing complete - 2 bad channels removed
  • Drift estimate: 15.3 um
  • Kilosort4 found 45 units
  • Quality metrics computed
  • Allen curation: 28 good units, 12 MUA, 5 noise

正在使用「neuropixels-analysis」。 Check for drift and motion in my recording

預期結果:

  • Motion estimate: 12.8 um peak-to-peak
  • No severe drift detected
  • Nonrigid motion correction applied
  • Corrected recording saved to motion/corrected/

安全審計

安全
v4 • 1/17/2026

All 703 static findings are false positives. The scanner incorrectly flags markdown code block backticks as shell commands, scientific terminology (channel, detect, universal) as C2/crypto keywords, and documentation URLs as hardcoded URLs. This is a legitimate neuroscience analysis toolkit using SpikeInterface and Kilosort4 for scientific research.

19
已掃描檔案
5,689
分析行數
3
發現項
4
審計總數
審計者: claude 查看審計歷史 →

品質評分

82
架構
100
可維護性
87
內容
23
社群
100
安全
91
規範符合性

你能建構什麼

自動化尖峰排序流程

處理 Neuropixels 記錄,從原始數據到排序的尖峰,並計算品質指標以供發表使用。

批次處理工作流程

對多個記錄會話應用標準化的預處理和精選,以確保一致性。

與 Phy 整合

將排序後的數據導出至 Phy,以便手動審查和微調尖峰排序結果。

試試這些提示

加載和預處理
Load a SpikeGLX recording from /path/to/data and apply standard preprocessing including highpass filtering at 400 Hz, phase shift correction, and common median reference.
執行尖峰排序
Run Kilosort4 spike sorting on the preprocessed recording and compute quality metrics including SNR, ISI violations ratio, and presence ratio.
運動校正
Check for drift in my Neuropixels recording and apply motion correction if the estimated drift exceeds 20 microns.
AI 輔助精選
Generate a summary plot for unit 15 showing waveforms and autocorrelogram, then analyze whether it appears to be a well-isolated single unit based on the visualization.

最佳實務

  • Always check drift before spike sorting - drift above 20 um significantly impacts quality
  • Use GPU for Kilosort4 for 10-50x faster processing compared to CPU alternatives
  • Save preprocessed data to avoid recomputing filtering steps on subsequent runs
  • Review uncertain units in Phy - automated curation provides starting points for manual refinement

避免

  • Skipping drift estimation before spike sorting can lead to poor unit isolation
  • Applying phase shift correction to Neuropixels 2.0 data (only needed for 1.0 probes)
  • Using default curation thresholds without considering your experimental requirements
  • Processing full recordings without testing on a subset first to verify pipeline

常見問題

此技能支援哪些格式?
SpikeGLX, Open Ephys, and NWB formats are fully supported.
尖峰排序需要 GPU 嗎?
Kilosort4 requires CUDA GPU. CPU alternatives include SpykingCircus2, Mountainsort5, or Tridesclous2.
有哪些精選方法可用?
Allen Institute (permissive), IBL (standard), and strict (conservative single-unit criteria).
如何手動檢查結果?
Export to Phy using the export_to_phy script, then run phy template-gui to open the GUI.
AI 可以幫助精選嗎?
Yes, Claude Code can visually analyze waveform plots and provide expert curation recommendations for uncertain units.
建議哪些預處理步驟?
Highpass filter at 300-400 Hz, phase shift for NP 1.0, bad channel removal, and common median reference.