Multimodal mAgnetic Resonance imaGIng in Cardiovascular Disease (MAGIC)

Cardiovascular disease (CVD) remains a leading cause of global mortality and disability. Accurate and early assessment of cardiac structure, function, and myocardial tissue characteristics is crucial for optimal clinical management. Cardiac Magnetic Resonance (CMR) has evolved from single morphological imaging into advanced multi-modal imaging. By integrating Cine, Late Gadolinium Enhancement (LGE), T1/T2 mapping, and 4D flow techniques, multi-modal CMR serves as a "gold standard" that provides a comprehensive macroscopic and cellular-level evaluation, including the identification of myocardial fibrosis, edema, and complex hemodynamic alterations.

Despite its clinical potential, systematic research comparing the diagnostic efficacy and prognostic value of multi-modal CMR features across a broad spectrum of cardiovascular diseases (such as ischemic heart disease, non-ischemic cardiomyopathy, and valvular diseases) is still lacking. Furthermore, optimization of scan efficiency, standardization of sequence acquisition protocols, and the development of contrast-free or AI-driven virtual imaging techniques (such as deep learning-based virtual native enhancement ) represent critical avenues to enhance clinical utility.

This prospective, observational registry study is designed to address this gap by establishing a large-scale, standardized CMR imaging database. Approximately 2,000 patients with clinically suspected or confirmed CVD will be consecutively enrolled. Following routine clinical care pathways, participants will undergo a "one-stop" multi-modal CMR examination using 3.0T MRI scanners.

To refine imaging protocols and validate novel synthetic imaging algorithms, a subset/sub-cohort analysis will specifically investigate the impacts of acquisition parameters-comparing performance between pre-contrast Cine vs. post-contrast Cine, as well as 3-slice vs. 9-slice T1 mapping protocols-on downstream machine learning/deep learning models for virtual native enhancement and tissue characterization.

The primary objectives are to: 1) systematically delineate the imaging feature spectrum across different CVD subtypes; 2) optimize multi-modal CMR acquisition protocols and validate deep learning models for virtual image generation (e.g., VNE); 3) assess the sensitivity of hemodynamic and tissue-characterization parameters (especially T1 mapping and extracellular volume [ECV]) in detecting early cardiac damage; and 4) explore the correlation between multi-modal CMR parameters and major adverse cardiovascular events (MACE) during the follow-up period. Ultimately, this study aims to provide robust, evidence-based support for precision diagnosis and risk stratification in cardiovascular medicine.