YongchengYAO/LIDC-IDRI-Lite

About

This is a preprocessed redistribution of LIDC-IDRI (TCIA), which is released under the CC BY 3.0 license.

Dataset summary: 1013 chest CT scans with consensus lung-nodule segmentation masks derived from 4-radiologist contours.

Contents of this repository:

• Images/ — 1013 files
• Masks/ — 1013 files

📝 Landmark annotations, visualization figures and the benchmark plan files live in 🔥MedVision🔥, where you can load the complete images and annotations from dataset configs.

Relation to the source dataset

In the source	1308 series: 1018 CT, 237 DX, 53 CR
Excluded here	the 237 DX and 53 CR projection radiographs, and 5 CT series with duplicate-z slices
In this repo	1013 Images + 1013 Masks

Two exclusions reduce the collection to 1013 scans.

1. Non-CT series (1018 CT kept). The 237 DX and 53 CR (projection X-ray) series in the same TCIA collection are 2D radiographs, not volumes, and carry no nodule contours.
2. 5 duplicate-z CT series excluded (1018 → 1013): LIDC-IDRI-0085, -0146, -0418, -0572, -0979. Each contains two or more DICOM slices sharing one ImagePositionPatient z, so the series has no single well-defined volume. Because the image and the mask are reconstructed by different libraries — SimpleITK for the volume, pylidc for the annotation, which returns array indices into its own view — the two must agree index-for-index. Rather than depend on reproducing pylidc's internal tie-break, these series are dropped.

Why -Lite? The suffix marks this as a derived redistribution rather than a copy of the source. These are preprocessed volumes — every case has been format-converted, geometry-normalised and reoriented to RAS+ — and for some sources cases or modalities are excluded as well (see the table above). Use it to reproduce MedVision, not as a substitute for the original release. See Preprocessing below for exactly what was changed.

Preprocessing

• DICOM CT series converted to nii.gz and standardized to RAS+ orientation.

• Masks are consensus binary nodule masks built from the four radiologists' XML contours via pylidc (50% consensus level).

• Series containing duplicate-z slices are excluded rather than de-duplicated, so no shipped case depends on two libraries pruning a volume identically.

• 880 of the 1013 scans contain at least one nodule; the remainder carry an all-zero mask.

• 8 patients contributed 2 CT series each, so every series gets a unique case ID.

Segmentation Labels

labels_map = {
    "1": "lung nodule"
}

Landmarks

landmarks_map = {
    "P1": "most right/anterior/superior endpoint of the major axis",
    "P2": "most left/superior/inferior endpoint of the major axis",
    "P3": "most right/anterior/superior endpoint of the minor axis",
    "P4": "most left/superior/inferior endpoint of the minor axis"
}

News

• [25 Jul, 2026] Initial release. This dataset is integrated into 🔥MedVision🔥, where you can use these config names to load data in python:

  • LIDC-IDRI_BoxSize_Task01_Axial_Test
  • LIDC-IDRI_BoxSize_Task01_Axial_Train
  • LIDC-IDRI_BoxSize_Task01_Coronal_Test
  • LIDC-IDRI_BoxSize_Task01_Coronal_Train
  • LIDC-IDRI_BoxSize_Task01_Sagittal_Test
  • LIDC-IDRI_BoxSize_Task01_Sagittal_Train
  • LIDC-IDRI_MaskSize_Task01_Axial_Test
  • LIDC-IDRI_MaskSize_Task01_Axial_Train
  • LIDC-IDRI_MaskSize_Task01_Coronal_Test
  • LIDC-IDRI_MaskSize_Task01_Coronal_Train
  • LIDC-IDRI_MaskSize_Task01_Sagittal_Test
  • LIDC-IDRI_MaskSize_Task01_Sagittal_Train
  • LIDC-IDRI_TumorLesionSize_Task01_Axial_Test
  • LIDC-IDRI_TumorLesionSize_Task01_Axial_Train

Data Usage Agreement

By using the dataset, you agree to the terms as follow.

• You must comply with the original CC BY 3.0 license terms of the source dataset.
• You are recommended to refer to the source of this dataset in any publication: https://huggingface.co/datasets/YongchengYAO/LIDC-IDRI-Lite
• You must cite the original publication(s):
  • https://doi.org/10.1118/1.3528204

Official Release

For more information, please go to the official site: https://www.cancerimagingarchive.net/collection/lidc-idri/

Download from Huggingface

# python
from huggingface_hub import snapshot_download
snapshot_download(repo_id="YongchengYAO/LIDC-IDRI-Lite", repo_type='dataset', local_dir="/your/local/folder")