Datasets:

nahidhasan
/

radgenome-ct-reshaped-tiny

license: mit
task_categories:
  - image-segmentation
  - image-classification
tags:
  - medical
  - ct-scan
  - radiology
  - chest-ct
size_categories:
  - n<1K
configs:
  - config_name: default
    data_files:
      - split: train
        path: train/*.parquet

RadGenome ChestCT Reshaped Tiny Dataset

This dataset contains resized chest CT scans from the RadGenome-ChestCT dataset.

Dataset Details

Original Resolution: 900x900xN
Resized Resolution: 300x300xN
Format: NIfTI (.nii.gz)
Number of Volumes: 253
Space Reduction: ~89% (resized to 1/9th of original spatial dimensions)

Dataset Structure

Each entry contains:

volumename: Name of the CT volume file (string)
anatomy: Anatomical region information (string)
sentence: Associated radiology report sentence (string)
volume_path: Relative path to the .nii.gz file (string)

Columns

Column	Type	Description
volumename	string	CT volume filename
anatomy	string	Anatomical region
sentence	string	Radiology report text
volume_path	string	Path to .nii.gz file

Usage

from datasets import load_dataset
import nibabel as nib

# Load the dataset
ds = load_dataset("nahidhasan/radgenome-ct-reshaped-tiny")

# Access dataset information
print(f"Number of samples: {len(ds['train'])}")

# Access a single entry
sample = ds['train'][0]
print(f"Volumename: {sample['volumename']}")
print(f"Anatomy: {sample['anatomy']}")
print(f"Sentence: {sample['sentence']}")
print(f"Volume path: {sample['volume_path']}")

# Download and load the NIfTI file
# nii = nib.load(sample['volume_path'])
# data = nii.get_fdata()
# print(f"Volume shape: {data.shape}")

Source

This is a resized subset of the RadGenome-ChestCT dataset.

Processing

CT volumes were resized from 900x900xN to 300x300xN using trilinear interpolation while preserving the slice dimension.