Skip to content

Core Pipeline

The core pipeline modules handle the main spike sorting workflow from raw data to sorted units.

Pipeline Modules

blech_exp_info.py

Pre-clustering step to annotate channels and save experimental parameters.

Key Functions:

  • Electrode layout configuration
  • Digital input selection
  • Taste/stimulus naming
  • Palatability ranking
  • Laser configuration

Usage:

python blech_exp_info.py /path/to/data

blech_init.py

Initialize directories and prepare data for clustering.

Key Functions:

  • Directory structure creation
  • Data file organization
  • Initial parameter setup

Usage:

python blech_init.py

blech_common_avg_reference.py

Perform common average referencing on electrode data.

Key Functions:

  • CAR group processing
  • Signal referencing
  • Artifact reduction

Usage:

python blech_common_avg_reference.py

blech_process.py

Core spike extraction and clustering module.

Key Functions:

  • Spike detection
  • Feature extraction
  • Clustering algorithms
  • UMAP dimensionality reduction

Spike Detection:

By default, spike detection uses a rolling (adaptive) threshold that computes the noise level independently for each time window. This allows detection to adapt to local noise variations across the recording.

  • use_rolling_threshold: Enable/disable rolling threshold (default: true)
  • rolling_threshold_window: Window size in seconds (default: 5.0)
  • rolling_threshold_step: Step size in seconds (default: 5.0)
  • waveform_threshold: Threshold multiplier in standard deviations (default: 5)

When rolling threshold is enabled, the threshold is computed as waveform_threshold * MAD / 0.6745 for each window, where MAD is the Median Absolute Deviation. This is more robust to outliers than standard deviation.

Set use_rolling_threshold: false in sorting_params.json to use a single global threshold computed over the entire recording.

Example rolling threshold plots in the migration guide

Usage:

# Usually called via blech_run_process.sh for parallel execution
python blech_process.py <electrode_number>

blech_post_process.py

Add selected units to HDF5 file after manual curation.

Key Functions:

  • Unit selection
  • HDF5 file updates
  • Metadata management

Usage:

python blech_post_process.py

blech_units_plot.py

Plot waveforms of selected spikes for visualization.

Key Functions:

  • Waveform plotting
  • Unit visualization
  • Quality metrics display

Usage:

python blech_units_plot.py

blech_make_arrays.py

Generate spike-train arrays for analysis.

Key Functions:

  • Spike train generation
  • Trial alignment
  • Array formatting

Usage:

python blech_make_arrays.py

Pipeline Flow

Raw Data
    ↓
blech_exp_info.py (Setup)
    ↓
blech_init.py (Initialization)
    ↓
blech_common_avg_reference.py (Referencing)
    ↓
blech_run_process.sh (Parallel Processing)
    ↓
blech_post_process.py (Unit Selection)
    ↓
blech_units_plot.py (Visualization)
    ↓
blech_make_arrays.py (Array Generation)
    ↓
Sorted Units

Configuration Files

The pipeline uses several JSON configuration files:

  • sorting_params.json - Clustering parameters
  • spike_detection_params.json - Detection thresholds
  • waveform_classifier_params.json - Classifier settings

See Also