Datasets:

AmirEhsan1995
/

lodopab-ct-glimpse

license: odc-by
pretty_name: LoDoPaB-CT (GLIMPSE subsets)
tags:
  - computed-tomography
  - medical-imaging
  - inverse-problems
task_categories:
  - image-to-image

LoDoPaB-CT subsets for GLIMPSE

Processed image subsets used by GLIMPSE (paper). Sinograms are not stored; they are rendered on the fly by the GLIMPSE data pipeline (ODL / scikit-image).

Layout

Split	Contents	Format
`train/`	LoDoPaB-CT training slices	`.npy` float32 arrays
`test/`	LoDoPaB-CT test slices	`.npy` float32 arrays
`ood/`	out-of-distribution brain images	`.jpg`

import numpy as np  # a .npy slice
img = np.load("train/0.npy")        # (H, W) float32 in [0, 1]

License & attribution

This repo mixes two sources with different (but both attribution-only) licenses; credit both when reusing.

train/ + test/ — LoDoPaB-CT (Leuschner et al., Scientific Data 2021), ODC-By v1.0, DOI 10.5281/zenodo.3384092. Built on LIDC-IDRI from TCIA (CC BY 3.0). Slices resized/curated for GLIMPSE; originals at the Zenodo DOI.

ood/ — CT-ICH intracranial-hemorrhage scans (Hssayeni et al., Data 2020), CC BY 4.0, PhysioNet ct-ich, DOI 10.13026/4nae-zg36.

Tip: for clean license tagging you may prefer two separate HF dataset repos (LoDoPaB subset odc-by; CT-ICH OOD cc-by-4.0).

@article{leuschner2021lodopabct,
  title   = {LoDoPaB-CT, a benchmark dataset for low-dose computed tomography reconstruction},
  author  = {Leuschner, Johannes and Schmidt, Maximilian and Baguer, Daniel Otero and Maass, Peter},
  journal = {Scientific Data}, volume = {8}, number = {1}, pages = {109}, year = {2021}
}
@article{hssayeni2020ctich,
  title   = {Intracranial Hemorrhage Segmentation Using a Deep Convolutional Model},
  author  = {Hssayeni, Murtadha and Croock, Muayad and Salman, Aymen and Al-khafaji, Hassan and Yahya, Zakaria and Ghoraani, Behnaz},
  journal = {Data}, volume = {5}, number = {1}, pages = {14}, year = {2020}
}