Skip to content

README.md

Latest commit

 

History

History
88 lines (71 loc) · 3.99 KB

README.md

File metadata and controls

88 lines (71 loc) · 3.99 KB

Automated Cardiac Diagnosis Challenge (ACDC)

The Automated Cardiac Diagnosis Challenge (ACDC) dataset consists of 100 training and 50 testing cine-MRI images, distributed among 5 different pathologies (see section III of the paper for more details): normal subjects (NOR), myocardial infarction (MINF), dilated cardiomyopathy (DCM), hypertrophic cardiomyopathy (HCM), and abnormal right ventricle (ARV). The data was acquired from one hospital in France over 6 years using two different scanners/. The segmentaiton labels of LV, RV, and MYO are provided for ED and ES phases. The volume and EF can thus be derived. Images are stored in NIfTI format with 12-35 time frames per patient. After resampling to 1mm x 1mm x 10mm, images have 5-11 slices.

Pathologies and criteria
Abbreviation Name Criteria
NOR Normal/healthy subjects LV EF >50%; LV EDV <90mL/m2 for men and <80mL/m2 for women; wall thickness in diastole <12mm; RVV <100mL/m2; RV EF >40%
MINF Myocardial infarction LV EF <40%; and abnormal myocardial infarction.
DCM Dilated cardiomyopathy LV EF <40%; LV >100mL/m2; wall thickness in diastole <12mm
HCM Hypertrophic cardiomyopathy LV EF >55% (normal); wall thickness in diastole >15mm
ARV Abnormal right ventricle RV >110mL/m2 for men and >100ml/m2 for women; RV EF <40%

Download Dataset

Note

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

For MacOS users, the wget command may not work. Please download the dataset manually and unzip it. Or install brew and run brew install wget.

The dataset can be downloaded from https://www.creatis.insa-lyon.fr/Challenge/acdc/.

wget -O acdc.zip https://humanheart-project.creatis.insa-lyon.fr/database/api/v1/folder/637218e573e9f0047faa00fc/download
UNZIP_DISABLE_ZIPBOMB_DETECTION=TRUE unzip acdc.zip

After unzipping the file, the files will have the following structure:

acdc.zip
database/
├── MANDATORY_CITATION.md
├── training/
│   ├── patient001/
│   │   ├── Info.cfg
│   │   ├── MANDATORY_CITATION.md
│   │   ├── patient001_4d.nii.gz
│   │   ├── patient001_frame01.nii.gz
│   │   ├── patient001_frame01_gt.nii.gz
│   │   ├── patient001_frame12.nii.gz
│   │   ├── patient001_frame12_gt.nii.gz
│   ├── patient002/
│   ├── ...
├── testing/
│   ├── patient101/
│   ├── patient102/
│   ├── ...

The Info.cfg file contains the group, height, weight, number of frames, and the frame number of end-diastole and end-systole for for each patient. For example, the file for patient 1 contains the following information:

ED: 1
ES: 12
Group: DCM
Height: 184.0
NbFrame: 30
Weight: 95.0

Preprocessing

The preprocessing is performed on ED and ES images with the following steps:

  • resampling the ED/ES SAX images to 1.0 x 1.0 x 10.0 mm.
  • cropping the images to 192 x 192 for xy, based on the LV center at ED frame.
  • normalizing the values to [0, 1].
acdc_preprocess

The CLI requires to install the cinema package at the root of repository with pip install -e ..