Skip to content

Sample Data and Scripts

This page describes the sample data and scripts available in examples/data/sample_data/ for testing and demonstrating PySPIM functionality.

Contents

Jupyter Notebooks

  • basic_usage.ipynb: Main example notebook demonstrating the complete PySPIM workflow
  • Data loading with uManager acquisition
  • ROI detection and cropping
  • Deskewing for both heads
  • Registration and transformation
  • Deconvolution

SLURM Job Scripts

The sh/ directory contains SLURM job scripts for batch processing:

  • 1_deskew.sh: CPU-based deskewing
  • 2_register.sh: GPU-based registration
  • 3_transform.sh: Data transformation
  • 4_deconvolve.sh: GPU-based deconvolution
  • 5a_rotate.sh: Data rotation

Usage

Running the Notebook

  1. Install dependencies: 

    uv pip install jupyter notebook matplotlib

  2. Start Jupyter: 

    cd examples/data/sample_data
jupyter notebook

  3. Open the notebook:

  4. Navigate to basic_usage.ipynb
  5. Run cells sequentially

Running SLURM Jobs

  1. Modify paths in the shell scripts to point to your data
  2. Submit jobs: 
    sbatch sh/1_deskew.sh
sbatch sh/2_register.sh
sbatch sh/3_transform.sh
sbatch sh/4_deconvolve.sh
sbatch sh/5a_rotate.sh

Data Requirements

Input Data Format

The scripts expect: - uManager acquisition format: Micro-Manager data files - TIFF files: Standard TIFF format - HDF5 files: Hierarchical data format

PSF Files

For deconvolution, you need: - PSF-A: Point spread function for head A (.npy format) - PSF-B: Point spread function for head B (.npy format)

Acquisition Parameters

Standard diSPIM parameters: - Step size: 0.5 microns - Pixel size: 0.1625 microns - Angle: Ī€/4 radians (45 degrees) - Camera offset: 100 (PCO.edge)

Customization

Using Your Own Data

  1. Update paths in the notebook: 

    root_fldr = "/path/to/your/data"
acq = "your_acquisition_name"

  2. Modify parameters: 

    step_size = 0.3      # Your step size
pixel_size = 0.1     # Your pixel size
theta = math.pi / 6  # Your angle

  3. Load your PSFs: 

    psf_a = numpy.load("/path/to/your/psf_a.npy")
psf_b = numpy.load("/path/to/your/psf_b.npy")

SLURM Configuration

Adjust SLURM parameters in the shell scripts:

CPU Jobs: 

#SBATCH --cpus-per-task=4
#SBATCH --mem-per-cpu=50G

GPU Jobs: 

#SBATCH --gres=gpu:1
#SBATCH --constraint=gpu80
#SBATCH --mem-per-cpu=200G

Performance Tips

GPU vs CPU

  • Use GPU for: Registration, deconvolution, large datasets
  • Use CPU for: Deskewing, small datasets, development

Memory Management

  • Monitor memory usage during processing
  • Use chunked processing for large datasets
  • Adjust SLURM memory allocation as needed

Processing Pipeline

  1. Deskew (CPU): Correct light sheet artifacts
  2. Register (GPU): Align multi-view data
  3. Transform (CPU): Apply transformations
  4. Deconvolve (GPU): Restore image quality
  5. Rotate (CPU): Final orientation

Troubleshooting

Common Issues

  1. Memory Errors
  2. Reduce data size for testing
  3. Increase SLURM memory allocation

  4. Use CPU processing for large datasets

  5. GPU Errors

  6. Check CUDA installation
  7. Verify GPU memory availability

  8. Use CPU fallback if needed

  9. Import Errors

  10. Ensure PySPIM is installed: uv pip install -e packages/pyspim
  11. Check Python environment

Getting Help

File Structure

examples/data/sample_data/
├── basic_usage.ipynb          # Main workflow notebook
├── sh/                        # SLURM job scripts
│   ├── 1_deskew.sh           # CPU deskewing
│   ├── 2_register.sh         # GPU registration
│   ├── 3_transform.sh        # Data transformation
│   ├── 4_deconvolve.sh       # GPU deconvolution
│   └── 5a_rotate.sh          # Data rotation
└── README.md                  # Usage instructions

Next Steps