CARE-Left Atrium

Left Atrial Segmentation and Analysis

Registration

To access the dataset, please register here.

Motivation

Figure 1.

Atrial fibrillation (AF), the most prevalent cardiac arrhythmia, is poised to escalate in frequency due to aging populations . Radiofrequency catheter ablation, a common AF therapy, faces challenges due to high recurrence rates . Cardiac digital twin provides personalized in-silico cardiac representations to infer multi-scale properties associated with cardiac mechanisms . It has shown great promise in personalized targeted ablation of persistent AF (see Fig. 1). To create a digital twin, it is important to reconstruct the left atrial (LA) geometry with the location of scars from LGE MRI . However, automatic quantification and analysis of LA scars can be quite challenging due to the low image quality, thin wall, the surrounding enhanced regions, and the complex patterns of scars . Deep learning (DL) methods have shown promise in LGE MRI analysis, yet their performance often falters in new domains due to domain shifts . CARE-Left Atrium aims to address these issues, driving the advancement of DL models that precisely delineate LA cavity and scars and ultimately revolutionize personalized AF treatment.

Task

Figure 2.

The target of this track is to automatically segment LA cavity and quantify LA scars from LGE MRI (see Fig. 2). The track will provide 500+ LGE MRIs and CTs globally, i.e., from multiple imaging centers around the world, for developing novel algorithms that can quantify or segment LA cavity and scars. The track presents an open and fair platform for various research groups to test and validate their methods on these datasets acquired from the clinical environment. To ensure data privacy, the platform will enable remote training and testing on the dataset from different centers in local and the dataset can keep invisible.

The selected papers will be published in our proceedings (see previous proceedings).

Topics may cover (not exclusively):

Data

Data acquisition information

We include 200+ multi-center LGE MRIs and 300 CTs (enhanced.nii.gz) from different countries, with manual segmentation of LA cavity (atriumSegImgMO.nii.gz) and/ or scarring region (scarSegImgM.nii.gz). All these clinical data have got institutional ethic approval and have been anonymized (please follow the data usage agreement, i.e., CC BY NC ND). The details of these LGE MRI are listed below:

Center A: 154 LGE MRIs

This data was original collected from Utah NAMIC-CARMA with permission for release. 2018 Atrial Segmentation Challenge refined the LA segmentation of Utah NAMIC-CARMA dataset before final release. Therefore, we adopted the refine dataset, and further fixed the resolution irregularities existing in this dataset. The clinical images were acquired with Siemens Avanto 1.5T or Vario 3T using free-breathing (FB) with navigator-gating. The spatial resolution of the 3D LGE MRI scan was 0.625 × 0.625 × 2.5 mm. The patient underwent an MR examination prior to ablation or was 3-6 months after ablation.

Center B: 20 LGE MRIs

This data was original collected from Beth Israel Deaconess Medical Center and was used in ISBI2012 Left Atrium Fibrosis and Scar Segmentation Challenge. We selected part of the dataset from this challenge and refine their manual segmentation before release. The clinical images were acquired with Philips Acheiva 1.5T using FB and navigator-gating with fat suppression. The spatial resolution of one 3D LGE MRI scan was 1.4 × 1.4 × 1.4 mm. The patient underwent an MR examination prior to ablation or was 1 month after ablation.

Center C: 20 LGE MRIs

This data was original collected from King’s College London and was used in ISBI2012 Left Atrium Fibrosis and Scar Segmentation Challenge. We selected part of the dataset from this challenge and refine their manual segmentation before release. The clinical images were also acquired with Philips Acheiva 1.5T using FB and navigator-gating with fat suppression. The spatial resolution of one 3D LGE MRI scan was 1.3 × 1.3 × 4.0 mm. The patient underwent an MR examination prior to ablation or was 3-6 months after ablation.

Center D: 300 CTs

This data was original collected from Fuzhou University Affiliated Provincial Hospital. We selected part of the dataset from this challenge and refine their manual segmentation before release.The clinical CT images were acquired with Siemens Force. The data were acquired at a resolution various from (0.30 x 0.30 x 0.5) to (0.80 x 0.80 x 0.5) mm.

Data split

The dataset has been divided into three main parts: training, validation, and test sets:

Task 1: LA scar quantification (MRI):

Task 2: LA cavity segmentation (MRI):

Task 3: LA multi-structure segmentation (CT):

Data Format

Each LGE MRI, CT and gold standard label(s) of patients will be provided in the NIfTI format as follows:

The submitted format of the prediction for the participants could be named as follows:

Metrics

The performance of LA cavity segmentation, LA scar quantification and cardiac structure segmentation results will be evaluated by：

Task 1: LA scar quantification (MRI):

Task 2: LA cavity segmentation (MRI):

Task 3: LA multi-structure segmentation (CT):

Rules

  1. External data sets and pre-trained models are allowed in this track.
  2. Only automatic methods are permitted.
  3. Participants are encouraged to attempt all tasks, but they also can choose to focus on one of them.
  4. We will only award prizes for LA scar quantification (MRI) and LA multi-structure segmentation (CT).

Leaderboards

Leaderboards will be released after test results submission.

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{zhuang2016multi,
  title={Multi-scale patch and multi-modality atlases for whole heart segmentation of MRI},
  author={Zhuang, Xiahai and Shen, Juan},
  journal={Medical image analysis},
  volume={31},
  pages={77--87},
  year={2016},
  publisher={Elsevier}
}

@article{li2022atrialjsqnet,
  title={AtrialJSQnet: a new framework for joint segmentation and quantification of left atrium and scars incorporating spatial and shape information},
  author={Li, Lei and Zimmer, Veronika A and Schnabel, Julia A and Zhuang, Xiahai},
  journal={Medical image analysis},
  volume={76},
  pages={102303},
  year={2022},
  publisher={Elsevier}
}

@article{GAO2023BayeSeg,
  title = {BayeSeg: Bayesian modeling for medical image segmentation with interpretable generalizability},
  journal = {Medical Image Analysis},
  volume = {89},
  pages = {102889},
  year = {2023},
  author = {Shangqi Gao and Hangqi Zhou and Yibo Gao and Xiahai Zhuang},
}

Contact

If you have any questions regarding the CARE-Whole Heart track, please feel free to contact: care26challenge@163.com or care2026challenge@outlook.com.