About Us

Welcome to the Cardiac MRI Reconstruction Challenge 2025 (CMRxRecon2025)!  

The CMRxRecon2025 (Towards Foundation Model) Challenge is a part of the 28th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2025, which will be held from September 23rd to 27th 2025 in Daejeon, Republic of Korea.

Learn more

Motivation

The objective of establishing the CMRx series challenges is to provide a benchmark that enables the research community to contribute to the work of accelerated CMR imaging with universal approaches that allow more diverse applications and better performance in real-world deployment in various environments. The previous CMRxRecon2023 and CMRxRecon2024 dataset did not cover multi-center, multi-vendor, and multiple diseases. Therefore, this year we aim to make an important leap towards real-world clinical scenarios.

Background

Cardiac MRI (CMR) has become an essential tool for diagnosing and evaluating cardiovascular diseases, offering multi-parametric, high-resolution anatomical and functional data. CMR reconstruction from highly under-sampled k-space data has gained significant attention in recent research. Numerous AI-based image reconstruction algorithms have shown potential in improving imaging performance and patient experience in recovering high-quality images from aggressively undersampled k-space measurements. However, the complexity and diversity of CMR scans in real-world applications, involving various image contrasts, sampling trajectories, equipment vendors, anatomical structures, and disease types, present a great challenge for existing AI-based reconstruction methods, which are usually developed for only one or a few specific scanning settings. In practice, there are often inevitable distribution mismatches between the training data and target data, due to the diversities listed above. Therefore, building and validating universal and robust reconstruction models for handling these diversities remain to be critical technical challenges for multi-parametric CMR imaging.

We aim to make an important leap towards real-world clinical scenarios by extending the challenge scope in two directions:

1) To evaluate the robustness and generalization performance of reconstruction foundation models in more than 5 centers and 10 scanners, all those data are unseen during training stage.

2) To evaluate the clinical performance of reconstruction foundation models under no less than 5 cardiovascular diseases (e.g., hypertrophic cardiomyopathy, dilated cardiomyopathy, myocardial infarction, coronary artery disease, arrhythmias), all those data are unseen during training stage.

Timeline

The schedule of the challenge is as follows. All deadlines are Pacific Standard Time (PST +11:59).