Description

In recent years, many machine/deep learning models have been proposed to accurately quantify and diagnose cardiac pathologies [1-4]. These models rely on accurate segmentation of the ventricles to allow the extraction of robust clinical features in practice. In particular, the segmentation of the right ventricle (RV) is a challenging task due to the highly complex and variable shape of the RV and its ill-defined borders in cardiac MR images [5]. Yet, quantitative features and clinical indices of right ventricular function are sensible to subtle changes in shape and texture [6]. Furthermore, when machine/deep learning models are tested on unseen datasets acquired from distinct disease groups or clinical centers, the segmentation accuracy can be greatly reduced [7]. Despite recent advances in deep learning, robust segmentation of the RV continues to pose challenges in practice and there is a need for new methods to handle the inherent geometrical and textural complexities, in particular in the presence of RV related pathologies (e.g. Dilated Right Ventricle, Tricuspid Insufficiency, Arrhythmogenesis, Tetralogy of Fallot and Interatrial Comunication). The last MICCAI challenge to focus on right ventricular segmentation took place in 2012, well before the deep learning revolution, and was based on 48 cases from a single clinical center. In this challenge, we invite participants to implement and evaluate advanced approaches based on machine/deep learning for right ventricular segmentation in a multi-disease, multi-view and multi-center setting. Specifically, we extend last year’s M&Ms challenge and dataset, which included mostly pathologies of the left ventricle, by presenting a large dataset of 450 cardiac MRI datasets which will includes pathologies that are associated with various right ventricular abnormalities and remodelling, including Dilated Right Ventricle, Tricuspid Insufficiency, Arrhythmogenesis and Tetralogy of Fallot and Interatrial Comunication. A novel aspect of this challenge is the inclusion of long-axis images to help the automatic definition of the basal plane of the RV, which can be confused with the right atrium. The challenge will be supported by the H2020 euCanSHare project (www.eucanshare.eu), which is building a multi-center big data platform for cardiovascular personalised medicine research. References [1] Bernard, Olivier, et al. "Deep learning techniques for automatic MRI cardiac multi-structures segmentation and diagnosis: Is the problem solved?." IEEE transactions on medical imaging 37.11 (2018): 2514-2525. [2] Tran, Phi Vu. "A fully convolutional neural network for cardiac segmentation in short-axis MRI." arXiv preprint arXiv:1604.00494 (2016). [3] Isensee, Fabian, et al. "Automatic cardiac disease assessment on cine-MRI via time-series segmentation and domain specific features." International workshop on statistical atlases and computational models of the heart. Springer, Cham, 2017. [4] Zotti, Clément, et al. "GridNet with automatic shape prior registration for automatic MRI cardiac segmentation." International Workshop on Statistical Atlases and Computational Models of the Heart. Springer, Cham, 2017. [5] Petitjean, Caroline, et al. "Right ventricle segmentation from cardiac MRI: a collation study." Medical image analysis 19.1 (2015): 187-202. [6] Gillies, Robert J., Paul E. Kinahan, and Hedvig Hricak. "Radiomics: images are more than pictures, they are data." Radiology 278.2 (2016): 563-577. [7] Zhuang, Xiahai et al. “Evaluation of algorithms for Multi-Modality Whole Heart Segmentation: An open-access grand challenge.” Medical Image Analysis (2019).
Date made available2 Mar 2021
PublisherZenodo

Cite this