GUI for Neuropixels preprocessing, curation, quality metrics, and post-processing.

• Add AP .bin files by drag/drop, file picker, or folder scan.
• Queue-based execution with live log/progress.
• Integrated ecephys_spike_sorting modules:
  • catGT_helper (with retry logic and failure gating)
  • ks4_helper / kilosort_helper
  • kilosort_postprocessing
  • noise_templates
  • mean_waveforms
  • quality_metrics
  • tPrime_helper
• Auto-generated input/output JSON per run.
• CatGT success is enforced before downstream ecephys spike sorting when CatGT is enabled.
• Built-in CatGT and TPrime string builders so users can generate common command fragments, including analog/digital and bit-field CatGT extractors, from structured inputs instead of typing raw flags.

• Launch phy template-gui params.py.
• Bombcell thresholds editor with:
  • live preview of labels
  • histogram view of selected metric + threshold lines
  • Apply settings + optional live update
  • auto quality-metrics computation if metrics.csv is missing
  • Run py_bombcell to compute BombCell metrics + plots (bombcell/templates._bc_qMetrics.csv)

• Run/re-run quality metrics.
• Optional Run py_bombcell backend for BombCell-native quality metrics and plots.
• Interactive table + filter + metric histogram.

• Auto-compute and auto-plot workflow (no separate compute buttons).
• Unit list with filters and Show good units only.
• Export selected dataset units to nested HDF5 (all units or good units only).
• Unit quality side panel (labels + key metrics).
• Unit basics:
  • raster for selected units
  • optional instantaneous firing-rate overlay
  • mean waveform with mean +/- SEM
• Raw explorer with settings-driven refresh and channel/depth Y-axis modes.
• Correlogram with histogram + alternate line-style cross-correlogram view.
• Condition PSTH and Network views auto-refresh on relevant changes.

• App-level controls:
  • Light/Dark plot theme
  • Grid on/off
• Applied across curation, quality metrics, and post-processing plots.

Core Python packages are listed in requirements.txt:

• PySide6
• pyqtgraph>=0.14.0,<0.15
• numpy
• pandas
• scipy
• matplotlib
• tqdm
• numba
• scikit-learn
• imbalanced-learn
• statsmodels
• networkx
• psutil
• joblib
• h5py
• seaborn
• cachecache
• upsetplot
• pyarrow
• ipython
• cmcrameri
• pillow

GPU-backed filtering/whitening in npyx also needs CuPy. For NVIDIA systems with a CUDA 12 driver/runtime, install:

• conda install -c conda-forge cupy cuda-version=12

Or, in an existing environment:

• python -m pip install cupy-cuda12x

Example with conda:

conda create -n neuropygui python=3.10 -y
conda activate neuropygui

Or from the bundled environment file:

conda env create -f environment.yml
conda activate neuropygui

From the project root (this folder):

pip install -r requirements.txt

If you are installing into an older Python 3.9 environment such as ks4_ece, keep in mind that pyqtgraph 0.14 requires Python 3.10+. The bundled environment.yml already targets Python 3.10 and is the recommended path for rebuilding the app or packaging a new .exe.

If you want full preprocessing/curation stack:

pip install phy

ecephys_spike_sorting and py_bombcell are bundled inside this app folder. No external source checkout is required.

External tool dependencies that must still be installed separately:

• CatGT
• TPrime
• C_Waves-win
• Kilosort4 (Python package/runtime and compatible environment)

Configure these paths in the Preprocessing tab:

• CatGT executable dir
• TPrime executable dir
• C_Waves executable dir
• KS4 repository dir
• Kilosort temp dir

From the project root (this folder):

python main.py

If you have multiple Python installations on PATH, prefer:

conda run -n ks4_ece python main.py

• Preprocessing expects SpikeGLX-style AP files (*.imecX.ap.bin).
• For curation previews, metrics.csv is loaded from the selected Kilosort folder.
• For post-processing quality columns and good-unit filtering, labels are read from:
  • bombcell_labels.csv (preferred)
  • cluster_group.tsv (fallback)