Fudan University / Imperial College London (Wang, Yang, Wang et al.)
Model weights not public. Contact creators for more information.
Generalist reconstruction foundation model for accelerating cardiac MRI (CMR) acquisition, designed to recover diagnostic-quality images from highly undersampled (8x-24x) multi-coil k-space data across heterogeneous scanners, field strengths, and cardiovascular diseases. Combines a CLIP-ViT-based module for semantic/contextual understanding of the anatomy being imaged with a physics-informed data-consistency reconstruction network, trained on MMCMR-427K -- the largest public multimodal CMR k-space database to date. Intended as an upstream substrate that feeds downstream segmentation, phenotyping, and diagnosis models (e.g. automated cardiac-phenotype extraction via nnU-Net) rather than replacing them. Released by the CMRxRecon-challenge consortium; code and the underlying database are public for academic, non-commercial use, but no separately downloadable pretrained checkpoint is provided.
Architecture
Hybrid
CLIP-ViT-based semantic/contextual encoder combined with a physics-informed, data-consistency-constrained unrolled reconstruction network for k-space to image mapping
Framework
PyTorch
Added to catalog
2026-08-10
427,465 multi-coil k-space acquisitions from 6,120 scans of 1,504 participants across 13 centers (4 public repositories + 9 clinical centers), 15 scanners (4 vendors, low-field to ultra-high-field), 12 CMR modalities, and 17 cardiovascular disease categories.
Reconstruction of fully-sampled cardiac MRI images from highly undersampled (8x-24x accelerated) multi-coil k-space data across diverse CMR modalities, scanners, and cardiovascular diseases