CARE-Whole Heart

Whole Heart Segmentation

Registration

Registration for CARE 2026 has been closed.

Motivation

Cardiovascular diseases (CVDs), as the leading cause of death globally, necessitate precise morphological and pathological quantification through segmentation of crucial cardiac structures from medical images. However, whole heart segmentation (WHS) faces challenges including heart shape variability during the cardiac cycle, clinical artifacts like motion and poor contrast-to-noise ratio, as well as domain shifts in multi-center data and the distinct modalities of CT and MRI. CARE-Whole Heart serves to inspire innovative solutions in the realms of biomedical imaging and computer vision, striving to overcome these challenges and advance automated WHS for enhanced understanding and treatment of CVDs.

Task

Figure 1. Overview of CARE-Whole Heart

The objective of this track is to achieve precise segmentation of seven substructures of the whole heart, with robustness against domain shifts (see Fig. 1).
The specific substructures, each associated with a unique label value, are:

  1. Left Ventricular Blood Cavity (LV) - Label value: 500
  2. Right Ventricular Blood Cavity (RV) - Label value: 600
  3. Left Atrial Blood Cavity (LA) - Label value: 420
  4. Right Atrial Blood Cavity (RA) - Label value: 550
  5. Myocardium of the Left Ventricle (Myo) - Label value: 205
  6. Ascending Aorta (AO) - Label value: 820; defined as the aortic trunk from the aortic valve to the superior level of the atria.
  7. Pulmonary Artery (PA) - Label value: 850; defined as the initial segment from the pulmonary valve to the bifurcation point.

Note on Great Vessels: The great vessels of interest, comprising the ascending aorta and pulmonary artery, are specifically defined due to variations in the fields of view across different scans. This uniform definition is crucial for ensuring consistency across evaluations. During the assessment, segmentation results for these vessels will be truncated to average lengths measured in healthy subjects, although participants are encouraged to extend their segmentation beyond these lengths. Our provided manual segmentations similarly cover more than the defined trunk measurements.

Data

About the data

1) Dataset overview: CARE-Whole Heart has 246 cases collected from 8 centers, supporting in-distribution and out-of-distribution performance evaluation for methods. Moreover, the diversity and clinical relevance of the dataset offers a broader spectrum of anatomical and pathological variations.

2) Scanner: Philips 64-slice CT scanner, SIEMENS SOMATOM Force CT scanner, GE Revolution_CT scanner, Philips Achieva 1.5T, Siemens Avanto 1.5T, Toshiba Aquilion ONE CT scanner.

3) Data format: The data are all in Nifty format. Each training case has one CT or MRI scan with its corresponding label.

Training data (106 Cases)

Center Num. patients Modalities
A 20 CT
B 20 CT
C & D 20 MRI
E 26 MRI
G 20 CT 
A ct_train
  |-- Case1001_image.nii.gz
  |-- Case1001_label.nii.gz
  |-- ...
B ct_train
  |-- Case2001_image.nii.gz
  |-- Case2001_label.nii.gz
  |-- ...
C and D mr_train
  |-- Case3001_image.nii.gz
  |-- Case3001_label.nii.gz
  |-- ...
...

Validation data (50 Cases)

Center Num. patients Modalities
A 20 CT
B 10 CT
C & D 20 MRI 
ct_val
|-- CaseCTVal001_image.nii.gz
|-- CaseCTVal002_image.nii.gz
|-- CaseCTVal003_image.nii.gz
|-- ...
mr_val
|-- CaseMRVal001_image.nii.gz
|-- CaseMRVal002_image.nii.gz
|-- CaseMRVal003_image.nii.gz
|-- ...

Test data (90 Cases)

Center Num. patients Modalities
A 20 CT
B 14 CT
C & D 20 MRI
F 16 MRI
H 20 CT 
ct_test
|-- CaseCTTest001_image.nii.gz
|-- CaseCTTest002_image.nii.gz
|-- CaseCTTest003_image.nii.gz
|-- ...
mr_test
|-- CaseMRTest001_image.nii.gz
|-- CaseMRTest002_image.nii.gz
|-- CaseMRTest003_image.nii.gz
|-- ...

Note on Validation and Test datasets: We have randomly shuffled the data from different centers and anonymized the center information to promote fairness.

Rules

Metrics

The performance of segmentation results will be assessed through:

Leaderboards

We will provide 2 leaderboards, representing stratified performance across different modalities, CT and MRI. In each leaderboard, in addition to DSC and HD of 7 cardiac substructures, we will also report DSC and HD of whole heart segmentation (DSC_WH and HD_WH), which are weighted average (according to their volumes) from 7 substructures.

Whole heart segmentation (CT)

Team DSCLV↑ DSCMyo↑ DSCRV↑ DSCLA↑ DSCRA↑ DSCAO↑ DSCPA↑ DSCWH↑ HDWH(mm)↓
BIT 0.9470 0.9307 0.9299 0.9617 0.9281 0.9609 0.8806 0.9414 11.3366
SJTUMedSeg 0.9456 0.9318 0.9292 0.9611 0.9273 0.9596 0.8787 0.9411 11.6841
VMHeart 0.9468 0.9334 0.9301 0.9624 0.9282 0.9603 0.8771 0.9420 11.7652
MEDII 0.9452 0.9285 0.9270 0.9601 0.9259 0.9584 0.8774 0.9396 11.7364
Monster 0.9431 0.9293 0.9288 0.9595 0.9258 0.9582 0.8716 0.9394 12.0265
AI_WHS_xjtlu 0.9421 0.9236 0.9203 0.9508 0.9161 0.9549 0.8735 0.9333 12.4595
M2Lab 0.9376 0.9271 0.9184 0.9595 0.8833 0.9475 0.8791 0.9305 12.8700
Slience 0.9337 0.9180 0.9196 0.9528 0.9119 0.9549 0.8637 0.9303 14.1676
NECS 0.9245 0.8817 0.8981 0.9331 0.9012 0.9340 0.8403 0.9091 13.3534
xigema 0.9161 0.8890 0.9012 0.9432 0.8778 0.9442 0.8531 0.9118 15.7054
TEAM1422_Whole_Heart 0.9197 0.9008 0.8966 0.9486 0.8885 0.9462 0.8346 0.9139 17.7390
DH 0.9354 0.9113 0.9131 0.9477 0.8614 0.9517 0.8709 0.9239 31.9159
CardiacEndeavours 0.8362 0.7327 0.8231 0.8766 0.8058 0.8278 0.6988 0.8097 18.2154
Kiruthiga.K 0.8362 0.7327 0.8231 0.8766 0.8058 0.8278 0.6988 0.8097 18.2154

Whole heart segmentation (MRI)

Team DSCLV↑ DSCMyo↑ DSCRV↑ DSCLA↑ DSCRA↑ DSCAO↑ DSCPA↑ DSCWH↑ HDWH(mm)↓
VMHeart 0.9406 0.8483 0.9036 0.9041 0.8930 0.9056 0.8476 0.9015 20.2336
BIT 0.9398 0.8497 0.9046 0.8986 0.8911 0.9020 0.8471 0.9001 20.3311
Monster 0.9398 0.8511 0.9038 0.9019 0.8904 0.9028 0.8412 0.9005 21.0346
AI_WHS_xjtlu 0.9379 0.8425 0.8967 0.8990 0.8895 0.8979 0.8401 0.8959 20.2396
MEDII 0.9390 0.8436 0.9024 0.8992 0.8907 0.9028 0.8470 0.8986 20.9197
SJTUMedSeg 0.9388 0.8443 0.9019 0.9013 0.8900 0.9049 0.8429 0.8990 21.1861
M2Lab 0.9375 0.8462 0.9033 0.8901 0.8867 0.9029 0.8404 0.8960 21.2334
DH 0.9256 0.8217 0.8614 0.8999 0.8817 0.8711 0.8019 0.8832 28.5251
Kiruthiga.K 0.8418 0.6665 0.7959 0.8296 0.8035 0.7829 0.6676 0.7829 22.9774
CardiacEndeavours 0.8418 0.6665 0.7959 0.8296 0.8035 0.7829 0.6676 0.7829 22.9774
Slience 0.9134 0.7707 0.8544 0.8774 0.8606 0.8271 0.8025 0.8608 43.9680
TEAM1422_Whole_Heart 0.7579 0.5580 0.6512 0.8000 0.7452 0.7817 0.4876 0.7064 43.8478
xigema 0.8613 0.7355 0.7895 0.7923 0.7611 0.8015 0.7338 0.7986 74.9196
NECS 0.8076 0.6866 0.5575 0.5578 0.3987 0.5224 0.0753 0.6165 73.5788

Citations

Please cite these papers when you use the data for publications:

@article{Zhuang2016MSMMA,
  Author = {Zhuang, Xiahai and Shen, Juan},
  Title = {Multi-scale patch and multi-modality atlases for whole heart
     segmentation of MRI},
  Journal = {Medical Image Analysis},
  Year = {2016},
  Volume = {31},
  Pages = {77-87},
}

@article{Zhuang2019MvMM,
  Author = {Zhuang, Xiahai},
  Title = {Multivariate Mixture Model for Myocardial Segmentation Combining
     Multi-Source Images},
  Journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
  Year = {2019},
  Volume = {41},
  Number = {12},
  Pages = {2933-2946},
}

@article{GAO2023BayeSeg,
  Author = {Gao, Shangqi and Zhou, Hangqi and Gao, Yibo and Zhuang, Xiahai},
  Title = {BayeSeg: Bayesian modeling for medical image segmentation with
     interpretable generalizability},
  Journal = {Medical Image Analysis},
  Year = {2023},
  Volume = {89},
  Pages = {102889},
}