paper-2026-opt-framework

Data and codes to reproduce publication results and figures

This README serves two purposes

1. Reproduce publication figures and data processing
2. Setup guide to the OPT framework

Paper reproducibility

Title: An unified open framework for Optical Projection Tomography: from acquisition to reconstruction DOI: XXX

Installation

This setup is using conda,

# get the repository
git clone https://github.com/QBioImaging/paper-2026-opt-framework.git

conda create -n paper2026-310 python==3.10
conda activate paper2026-310

# enter the repo
cd paper-2026-opt-framework

# install dependencies
pip install -r requirements.txt

# setup a kernel for the NBs
python -m ipykernel install --user --name "opt-framework"

further install tomopy

conda install --channel conda-forge tomopy

# or if you have cuda available
conda install --channel conda-forge tomopy "libtomo=*=cuda*"

Dependencies

The requirements contain standard python image processing libraries and packages developed by us. However for the purpose here, we import only specific modules from those, instead of using then as napari plugins, which aslo demonstrates the flexibility of the source code.

Data

raw data are deposited at Zenodo (should be bioimage? perhaps), here is programatic way to pull the data.

Download the raw_data.tar.gz and open the archive in the raw_data folder in this repository. That is why the raw_data folder is left empty.

Open archive is 10.4 GB

tar -xzvf raw_data.tar.gz -C raw_data/

Data folder structure

raw_data folder contains raw tiff files. processed_data contains corrected projections ready for the reconstruction, max size for 400 projections is 2.5 GB in .npy. data_output folder contains the fully sampled 3D recontructed volumes (9.5 GB .npy each), together with selected slices, which are for paper figure purposes.

Reconstruction volumes files are larger than the processed projections. Before reconstruction, the stack of projections is a size of 2048 x 1536 x 400, while after reconstruction for uniform resolution the volumes become 2048 x 1536 x 1536 meaning intrapolation from angles to camera line pixels. In other words the ratio in voxels corresponds to the ratio of file sizes: 9.5/2.5 = 1536/400.

Notebooks

First notebook 01_process_raw_data:

• shows a step by step processing pipeline, while saving the processed data to initially empty processed_data folder.
• produces Figure XXX of the paper demonstrating

Second notebook loads raw/processed data to reproduce figure of the paper.

OPT framework

Links to each tool for the setup is referenced in the article, with detailed instructions in each of the manuals. Here we provide the an ultrafast minimal setup.

Install ImSwitch OPT

conda create -n imswitchOTP ''

# install ImSwitch OPT
git clone ...

cd ...

pip install .

cd ..

Install napari and plugins

run napari in the imswitchOPT environment

python -m napari