Skip to content

Latest commit

 

History

History
79 lines (67 loc) · 3.07 KB

File metadata and controls

79 lines (67 loc) · 3.07 KB

Rescan

For access to the dataset, please contact Rhodri Davies (rhodri.davies@ucl.ac.uk).

The Rescan dataset was obtained from 5 United Kingdom institutions with 6 different MRI scanners of 2 field strengths (1.5T, 3T), 2 manufacturers (Siemens, Philips), and 3 models (Aera, Achieva, Avanto). The patients characteristics include healthy, myocardial infarction, left ventricular hypertrophy, cardiomyopathy, and other pathologies. For the same subject, two scans were acquired within a short time frame and thus the cardiac function is assumed to be the same. The objective is to evaluate machine learning models's variance in predicting EF, using coefficients of variation (CV) as the metric.

There are three sources of data

  • train: unpaired data with segmentation labels derived from a segmentation model trained separately
  • test: paired data with segmentation labels derived from a segmentation model trained separately
  • test_retest_100: paired data without segmentation labels, but with EDV/ESV values

Preprocess

Note

It is recommended to download the dataset to download the data under ~/.cache/cinema_datasets/rescan as the integration tests uses this path. Otherwise define the path using environment variable CINEMA_DATA_DIR.

The files should be stored in the following structure:

rescan/
├── pickle/
│   ├── train/
│   │   ├── A/
│   │   │   ├── s_0000/
│   │   │   │   ├── 2C.pickle
│   │   │   │   ├── 4C.pickle
│   │   │   │   ├── SAX.pickle
│   │   │   │   ├── SAX_segs.pickle
│   │   │   ├── s_0001/
│   │   │   ├── ...
│   │   ├── B/
│   │   ├── ...
│   ├── test/
│   │   ├── scan_00_A/
│   │   │   ├── 2C.pickle
│   │   │   ├── 4C.pickle
│   │   │   ├── SAX.pickle
│   │   │   ├── SAX_segs.pickle
│   │   ├── scan_00_B/
│   │   ├── ...
│   ├── test_retest_100/
│   │   ├── 1027/
│   │   │   │   ├── 2C.pickle
│   │   │   │   ├── 4C.pickle
│   │   │   │   ├── SAX.pickle
│   │   ├── 1153/
│   │   │   │   ├── 2C.pickle
│   │   │   │   ├── 4C.pickle
│   │   │   │   ├── SAX.pickle
│   │   ├── ...
├── processed/
│   ├── train/
│   │   ├── A/
│   │   │   ├── s_0000/
│   │   │   │   ├── lax_2c_t.nii.gz
│   │   │   │   ├── lax_4c_t.nii.gz
│   │   │   │   ├── sax_t.nii.gz
│   │   │   │   ├── sax_gt_t.nii.gz

The preprocessing is performed on SAX and LAX images with the following steps:

  • resampling the SAX images to 1.0 x 1.0 x 10.0 mm and LAX images to 1.0 x 1.0 mm.
  • cropping the images to SAX, LAX image intersection.
  • normalizing the values to [0, 1].
rescan_preprocess --data_dir pickle --out_dir processed