Artificial Intelligence in Cardiac Magnetic Resonance: Advances in Myocardial Tissue Characterization

Artificial Intelligence (AI) for Myocardial Tissue Characterization in Cardiac Magnetic Resonance (CMR) represents a rapidly maturing field with significant potential to revolutionize non-invasive cardiac diagnostics. Traditionally, CMR relies on gadolinium-based contrast agents (GBCA) to assess late gadolinium enhancement (LGE) for myocardial viability and structure. However, the use of GBCA poses risks, including nephrogenic systemic fibrosis and gadolinium deposition in tissues. Parametric native T1, T2 and T2* mapping, along with extracellular volume (ECV) quantification, have emerged as promising alternatives, offering quantitative pixel-wise myocardial tissue phenotyping. Despite their potential, the lack of standardized mapping sequences and thresholds for distinguishing between healthy and diseased myocardium has hindered their routine clinical application. Recent advancements in AI — including deep learning reconstruction, foundation models, generative approaches such as virtual or “synthetic” LGE, and radiomics-based tissue signatures — have opened new avenues for myocardial disease pattern recognition, leveraging existing myocardial signal features in non-contrast steady-state free precession cine imaging and T1 and T2 mapping. These AI-based solutions promise to mitigate the limitations of current methods, reduce reliance on contrast agents, shorten and even enable free-breathing acquisitions, and enhance diagnostic accuracy, particularly in resource-limited settings.This Research Topic aims to explore and develop AI strategies for myocardial tissue characterization in CMR, with the goal of advancing non-invasive diagnostic techniques. Specifically, the research seeks to address questions such as: How can AI improve the accuracy and reliability of myocardial tissue phenotyping? What are the most effective AI models for identifying myocardial disease patterns without contrast agents? How can multi-centre, multi-vendor generalisability and bias be addressed so that models perform reliably across diverse populations and scanners? What role can AI-derived imaging biomarkers play in risk stratification, prognosis, and guiding therapy in cardiomyopathies, ischaemic heart disease, myocarditis, and cardio-oncology? How can these AI solutions be standardized and integrated into clinical practice to benefit a broader patient population?We particularly welcome original research contributions, alongside reviews, methodological and validation studies, addressing — but not limited to — the following themes:• Development and validation of AI models for myocardial tissue phenotyping• Comparative studies of AI-based and traditional CMR techniques• Standardization of AI algorithms for clinical use and benchmarking across vendors, field strengths and centres• Impact of AI on reducing gadolinium use and scan times• AI applications in low-resource settings for cardiac diagnostics • Integration of AI with existing CMR workflows• Ethical considerations and regulatory challenges in AI-based CMR• Deep learning–based image reconstruction and accelerated, free-breathing mapping sequences• Generative AI for synthetic or “virtual” LGE and contrast-free fibrosis assessment• Foundation models, self-supervised learning, and federated learning for CMR tissue characterisation• Radiomics and AI-derived imaging biomarkers for prognosis and risk stratification• AI applications in specific disease contexts (e.g. cardiomyopathies, myocarditis, amyloidosis, cardio-oncology, ischaemic heart disease) • Explainability, uncertainty quantification, bias and fairness of AI models in CMR• Hybrid AI–CMR workflows combining tissue characterisation with strain, perfusion or 4D flow Topic editors: Jennifer Mancio, João Pedrosa, Stefan K Piechnik Source: https://www.frontiersin.org/research-topics/82023/artificial-intelligence-in-cardiac-magnetic-resonance-advances-in-myocardial-tissue-characterization