CARE-MyoPS

Myocardial Pathology Segmentation

Motivation

Figure 1. Myocardial pathology segmentation and its challenges. (A) Myocardial Pathology Segmentation: Scar and edema regions are marked in green and yellow, respectively. (B) Challenges of Myocardial Pathology Segmentation: The challenges include multi-center data, missing sequences, and misalignments in multi-sequence CMR images.

Myocardial infarction (MI) is a major cause of mortality and disability worldwide. Assessment of myocardial viability is essential in the diagnosis and treatment management of MI patients . Multi-sequence cardiac magnetic resonance (MS-CMR) images can provide valuable myocardial pathology information, which is important for the diagnosis and treatment management of patients. As shown in Figure 1 (A), balanced steady-state free precession (bSSFP) cine sequences present clear anatomical boundaries, while late gadolinium enhancement (LGE) and T2-weighted (T2) CMR sequences visualize myocardial scar and edema of MI, respectively.

Tasks

The target of this track is to segment myocardial pathology regions, specifically scar and edema, from multi-sequence CMR data. This track seeks innovative solutions to address MyoPS using real-world multi-sequence CMR data. We encourage participants to overcome challenges such as the inclusion of multi-center data, missing sequences for some centers , and misalignments in multi-sequence CMRs , as illustrated in Figure 1 (B).

The specific substructures, each associated with a unique label value, are:

1. Scar - Label value: 2221
2. Edema - Label value: 1220
3. Left ventricle - Label value: 500
4. Myocardium - Label value: 200
5. Right ventricle - Label value: 600

We will rank participant methods based on the settings (​Lb1–Lb4) detailed in the following table:

Leaderboard (Lb) for MyoPS track across targets and evaluation settings.​​ Lb1–Lb4 represent performance from different test centers and targets (e.g., Lb1 = scar segmentation peformance on center B). In-distribution refers to centers included in the training data, while out-of-distribution refers to unseen centers not used during training.

Target	In-distribution	Out-of-distribution
Scar	Lb1	Lb2
Edema	Lb3	Lb4

Data

Training data

Center	Num. patients	Sequences	Manual labels
A	81	LGE	Scar, left ventricle and myocardium
B	50	LGE, T2 and bSSFP	Scar, edema, left ventricle, myocardium and right ventricle
C	45	LGE, T2 and bSSFP	Scar, edema, left ventricle, myocardium and right ventricle
E	7	LGE and bSSFP	Scar, left ventricle, myocardium and right ventricle
F	9	LGE and bSSFP	Scar, left ventricle, myocardium and and right ventricle
G	8	LGE and bSSFP	Scar, left ventricle, myocardium and and right ventricle
H	35	LGE	Scar

Validation data

Center	Num. patients	Sequences	Manual labels
D	25	LGE, T2 and bSSFP	Scar, edema, left ventricle, myocardium

Test data

Center	Num. patients	Sequences	Manual labels
D	25	LGE, T2 and bSSFP	Scar, edema, left ventricle, myocardium and right ventricle
B	15	LGE, T2 and bSSFP	Scar, edema, left ventricle, myocardium and right ventricle

Metrics

The performance of scar and edema segmentation results will be evaluated by：

• Dice Similarity Coefficient (Dice)
• Precision (Pre)
• Sensitivity (Sen)
• Hausdorff Distance (HD in mm)

Rules

• Only automatic methods are acceptable. Participants must utilize algorithms that do not require manual intervention or human-assisted processes for the segmentation task.
• Pre-trained models are allowed in this track. The solutions could be developed with pre-trained fundation models, such as SAM, CLIP, and MedSAM.

Registration

To access the dataset, please register here.

Leaderboards

MyoPS Scar (ID)

Team	Scar_Dice↑	Scar_Sen↑	Scar_Pre↑	Scar_HD(mm)↓
JMT	0.4864	0.6580	0.4117	31.9086
ZXN	0.4329	0.5863	0.3833	37.4517
FireMan	0.4694	0.6572	0.3935	30.2447
king	0.4245	0.5416	0.3744	46.7791
marisabe	0.4841	0.6177	0.4221	29.2022
fpt	0.4933	0.6075	0.4390	28.7960
David	0.4248	0.5469	0.3772	46.8064
Space	0.4642	0.6750	0.3875	35.8338

MyoPS Scar (OOD)

Team	Scar_Dice↑	Scar_Sen↑	Scar_Pre↑	Scar_HD(mm)↓
JMT	0.6780	0.6126	0.7747	15.6920
ZXN	0.6667	0.6140	0.7549	16.1454
FireMan	0.6826	0.6394	0.7436	20.8305
king	0.7387	0.6888	0.8160	13.5691
marisabe	0.6573	0.5812	0.7775	17.2670
fpt	0.6459	0.5769	0.7550	24.1498
David	0.6654	0.6168	0.7747	16.4942
Space	0.5616	0.5991	0.5833	18.9891

MyoPS Edema (ID)

Team	Edema_Dice↑	Edema_Sen↑	Edema_Pre↑	Edema_HD(mm)↓
JMT	0.6061	0.6775	0.5873	32.1836
ZXN	0.6499	0.6430	0.6992	31.8488
FireMan	0.5935	0.7099	0.5361	31.1575
king	0.5187	0.6030	0.4826	46.2529
marisabe	0.5633	0.6430	0.5390	32.0530
fpt	0.6180	0.6247	0.6597	32.7830
David	0.5753	0.6227	0.5757	39.8774
Space	0.5953	0.6128	0.6310	33.8135

MyoPS Edema (OOD)

Team	Edema_Dice↑	Edema_Sen↑	Edema_Pre↑	Edema_HD(mm)↓
JMT	0.6655	0.5980	0.8039	21.8737
ZXN	0.6857	0.6339	0.7866	19.3169
FireMan	0.6819	0.6840	0.7211	22.2706
king	0.7585	0.7677	0.7715	14.7051
marisabe	0.6554	0.6371	0.7158	22.7204
fpt	0.6404	0.5548	0.7932	27.0451
David	0.6878	0.6208	0.8070	19.7881
Space	0.6392	0.5741	0.7551	22.4153

Citations

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

 @article{zhuang2019multivariate,
    title={Multivariate mixture model for myocardial segmentation combining multi-source images},
    author={Zhuang, Xiahai},
    journal={IEEE transactions on pattern analysis and machine intelligence},
    volume={41},
    number={12},
    pages={2933--2946},
    year={2019},
}

@article{qiu2023myops,
  title={MyoPS-Net: Myocardial pathology segmentation with flexible combination of multi-sequence CMR images},
  author={Qiu, Junyi and Li, Lei and Wang, Sihan and Zhang, Ke and Chen, Yinyin and Yang, Shan and Zhuang, Xiahai},
  journal={Medical image analysis},
  volume={84},
  pages={102694},
  year={2023},
}

 @article{ding2023aligning,
  title={Aligning multi-sequence CMR towards fully automated myocardial pathology segmentation},
  author={Ding, Wangbin and Li, Lei and Qiu, Junyi and Wang, Sihan and Huang, Liqin and Chen, Yinyin and Yang, Shan and Zhuang, Xiahai},
  journal={IEEE Transactions on Medical Imaging},
  year={2023},
}