Reinventing coronary MRI: Faster scans, clearer images
Coronary artery disease causes 50% of cardiovascular deaths. Current gold-standard imaging (X-ray angiography and CT) carries risks from invasiveness, radiation, and contrast agents. Coronary MR angiography (CMRA) is a safe, non-invasive alternative, but long scans, lower resolution, and motion sensitivity limit its use, creating a need for faster, higher-resolution, and motion-robust CMRA.
CardioMRI
Prof. René Botnar (King’s College London / Pontificia Universidad Católica de Chile), Alumnus Hans Fischer Senior Fellow | Ivan Kokhanovskyi (TUM University Hospital Rechts der Isar), Doctoral Candidate
Hosts: Prof. Marcus Makowski (TUM University Hospital Rechts der Isar), Prof. Daniel Rückert (TUM), Prof. Dimitrios Karampinos (TUM University Hospital Rechts der Isar)
Cardiovascular disease is the leading cause of death worldwide, and early detection of coronary artery disease and myocardial injury is essential for preventing serious events. Although X-ray angiography and CT provide excellent visualization of the coronary arteries, they require ionizing radiation, contrast agents, and – in the case of angiography – an invasive procedure, making them less suitable for repeated monitoring. Cardiovascular magnetic resonance (CMR) is a safer, non-invasive, radiation-free alternative, but conventional CMR is slow and complex, and it typically requires multiple scans to assess anatomy and tissue health. In this Fellowship we aimed to overcome these limitations by developing faster, more efficient, and more comprehensive MRI methods that could ultimately serve as practical alternatives to X-ray-based imaging.
The scientific concept was to combine advanced image acquisition with modern deep-learning reconstruction to dramatically accelerate 3-D cardiac imaging while improving image quality. A central goal was to enable rapid whole-heart scanning with high detail and without contrast agents. Initially, the Fellowship focused on creating a new “super-resolution” framework for coronary MR angiography (CMRA) that would enhance quickly acquired (one to two minutes), low-resolution data using deep learning, producing images with clarity approaching CT. As deep-learning methods matured, we expanded this approach to scar imaging and quantitative tissue mapping, transforming the Fellowship from improving a single MRI sequence into developing a unified strategy for assessing both cardiac anatomy and tissue properties in one scan.
Figure 1
2025 Phair et al. Reproduced with permission.
As part of dissemination efforts, we organized the CardioMRI Symposium in July 2023: From MR Physics and Novel Hardware to Artificial Intelligence: Making CMR a More Accessible and Affordable Imaging Modality. The meeting brought together national and international experts in cardiac MRI, computer science, engineering, and clinical practice, providing a platform to present early results, stimulate collaboration, and discuss how innovations developed in this Fellowship could help broaden global access to advanced cardiac imaging.
A major achievement was the creation of a highly accelerated, motion-corrected, super-resolution CMRA method [1] that reduced scan time fourfold and reconstruction time by 150 to 200 times. This technique enables submillimeter coronary imaging in predictable, clinically practical scan times while mitigating long acquisitions and blurring from cardiac and respiratory motion. These improvements bring CMRA closer to routine use for evaluating coronary artery disease in patients who require repeated imaging or cannot undergo CT.
Building on this success, we adapted the reconstruction framework to dark-blood late gadolinium enhancement (LGE) imaging, widely used for detecting myocardial scar. Conventional LGE requires multiple breath-held scans and is difficult for many patients. Using our accelerated, deep-learning-supported reconstruction, we produced high-quality 3-D LGE images [2] more quickly and with better visibility of subtle scars, which may improve diagnosis of prior infarction and identification of arrhythmogenic tissue
Figure 2
2025 Hajhosseiny et al. Reproduced with permission
Another important accomplishment was demonstrating simultaneous coronary lumen and high-risk plaque imaging using advanced non-contrast 3-D whole-heart CMR. We showed that both the bright-blood coronary lumen and vulnerable atherosclerotic plaque can be visualized in the same scan, allowing co-registered assessment without radiation or contrast agents [3]. This capability provides a powerful tool for risk stratification, longitudinal monitoring, and precision guidance of therapy.
A particularly impactful outcome was the development and application of the ACTION sequence [4], which aims to simplify the entire cardiac MRI examination. ACTION enables 3-D whole-heart imaging of bright-blood and black-blood anatomy together with joint T1/T2 tissue mapping – all from a single, efficient, free-breathing scan. Traditionally, these measurements require multiple 2-D breath-held scans with limited coverage. ACTION replaces this fragmented workflow with a unified 3-D acquisition that covers the entire heart, improving consistency, reducing exam time, and providing co-registered information on coronary anatomy, cardiac structure, and tissue properties. This approach has the potential to make comprehensive CMR more practical in a wide range of clinical settings.
Outcomes of the Fellowship will support wider clinical evaluation of these methods at multiple centers. The super-resolution CMRA and accelerated LGE techniques can be integrated with existing MRI systems, and ACTION provides a platform for future “all-in-one” cardiac MRI examinations. These innovations also align with efforts to expand MRI access by enabling high-quality imaging on lower-cost, low-field scanners, making advanced cardiac assessment feasible in resource-limited environments.
Figure 3
(T1 and T2 maps) for comprehensive myocardial tissue characterization such as e.g. fibrosis (T1) and inflammation (T2) detection.
2024 Kokhanovskyi et al. Reproduced with permission.
Looking forward, several promising directions arise from this work. One objective is to develop an even faster version of whole-heart CMRA [5], potentially achievable in a single breath-hold for compliant patients and in approximately one minute for free-breathing acquisitions. Another goal is to further refine ACTION and its reconstruction framework so that anatomy, coronary arteries, and tissue characteristics can be extracted more accurately and automatically using end-to-end deep-learning networks. Ultimately, we envision a fully integrated cardiac MRI exam in which a single rapid 3-D scan provides all the essential information currently obtained from multiple sequences, including anatomy, function, perfusion, and tissue characterization. Achieving this would transform cardiac MRI into a faster, simpler, and more widely accessible tool for diagnosing and monitoring heart disease.
To illustrate the versatility of these reactors, we investigated the selective photooxidation of tertiary alcohols at ambient conditions (1000 mbar, T = 23°C) in an anaerobic environment on bare TiO2 P25 upon UV irradiation [4]. Exemplified for 2-methyl-2-pentanol and 2-methyl-2-butanol, the reaction proceeds exclusively via the photo-induced, homolytic cleavage of the long alkyl chain, resulting in the respective ketone and the alkyl-moiety, which predominantly recombines with hydrogen upon alkane formation. Alternatively, the dimerization of two alkyl-radicals occurs as a side reaction and happens on bare TiO2 at enhanced alcohol surface concentrations. The reaction scheme is shown in Fig. 3. This demonstrates that alkyl-radical chemistry is enabled on bare TiO2, which is also relevant for other reactions, such as Kolbe-type ones.
These new reactor platforms open an avenue for fast and versatile testing of potential photocatalysts in the search for new processes with high selectivity that is not accessible by conventional thermal processes.
[1]
A. Phair et al (2025).
[2]
D. Si et al. (2026).
[3]
R. Hajhosseiny et al. (2025).
[4]
I. Kokhanovskyi et al. (2024).
[5]
A. Phair et al. (2024).
Selected publications
- A. Phair et al., “Super-MoCo-MoDL: A combined super-resolution and motion-corrected undersampled deep-learning reconstruction framework for 3D whole-heart cardiac MRI,” J. Cardiovasc. Magn. Reson., vol. 28, no. 1, art. no. 101990, 2025, doi: 10.1016/j.jcmr.2025.101990.
- D. Si et al., “Highly accelerated 3D water/fat LGE imaging with deep-learning motion estimation and motion corrected reconstruction,” In Proc. International Society of Magnetic Resonance in Medicine, 2026.
- R. Hajhosseiny et al., “Free-breathing, non-contrast, three-dimensional whole-heart coronary magnetic resonance imaging for the identification of culprit and vulnerable atherosclerotic plaque,” J. Cardiovasc. Magn. Reson., vol. 27, no. 1, art. no. 101898, 2025. doi: 10.1016/j.jocmr.2025.101898.
- I. Kokhanovskyi et al., C. Castillo-Passi, M. G. Crabb, C. Ganter, K. P. Kunze, R. Neji, D. Karampinos, M. R. Makowski, C. Prieto, and R. M. Botnar. “Free-breathing 3D whole-heart simultaneous bright- and black-blood anatomical imaging and T1/T2 mapping at 0.55T,” In Proc. International Society of Magnetic Resonance in Medicine, 2024, pp. 1487.
- A. Phair et al., “A motion-corrected deep-learning reconstruction framework for accelerating whole-heart magnetic resonance imaging in patients with congenital heart disease,” J. Cardiovasc. Magn. Reson., vol. 26, no. 1, art. no. 101039, 2024. doi: 10.1016/j.jocmr.2024.101039.