LAHARRAF/IntrinsicX-GAN

Reproducible pipeline for building the IntrinsicX dataset from five public sources: ChestX-ray14, DeepLesion, BraTS-Africa, LIDC-IDRI, and OpenBHB. It harmonizes 841,739 chest X-ray, CT, and brain MRI images into a standardized 256×256 format with patient-level train, validation, and test splits. # IntrinsicX Data Pipeline Reproducible pipeline that builds the **IntrinsicX** corpus from five public medical imaging archives — NIH ChestX-ray14, NIH DeepLesion, BraTS-Africa, LIDC-IDRI, and OpenBHB — harmonizing **841,739 images** into a single 256×256 HDF5 schema with patient-level 70/15/15 splits, and estimating healthy-tissue reference statistics for brain MRI and chest radiography. | Resource | Identifier | |---|---| | Dataset | doi.org | | Archived code | `doi.org]` | | Paper | *[]* | --- ## Scope **Contains** the code to regenerate the harmonized corpus and the reference statistics from the original public archives. **Does not contain** any imaging data — each source must be obtained directly under its own access terms. **Does not contain** the IntrinsicX-GAN model. The generative architecture that consumes this corpus is the subject of a separate publication and will be released alongside it. --- ## Requirements Python 3.10. Developed and executed in Google Colab with data staged on Google Drive. ```bash pip install -r requirements.txt ``` Approximately **140 GB** of free storage is required for the full corpus including native volumetric data; the harmonized 2D corpus alone occupies 25.49 GB. --- ## Source access | Source | Modality | Access requirement | |---|---|---| | NIH ChestX-ray14 | Chest X-ray | Open, no registration | | NIH DeepLesion | CT | Open, no registration | | BraTS-Africa | Brain MRI | TCIA account + data use agreement | | LIDC-IDRI | CT | TCIA, open | | OpenBHB | Brain MRI | HuggingFace account + access agreement | Supply credentials through environment variables (`HF_TOKEN`, TCIA credentials). **Never commit tokens or API keys.** --- ## Run order | # | Notebook | Produces | |---|---|---| | 1 | `download_NIH Chest X-ray Dataset.ipynb` | Harmonized chest X-ray HDF5 | | 2 | `download_deeplesion_dataset.ipynb` | Harmonized DeepLesion HDF5 | | 3 | `download_MRI_The …