Jonas Kleineisel¹, Philipp Eirich^1,2, Julius F. Heidenreich¹, Herbert Köstler¹, Thorsten A. Bley¹, and Tobias Wech^1
1Department of Diagnostic and Interventional Radiology, University of Würzburg, Würzburg, Germany, ²Comprehensive Heart Failure Center Würzburg, Würzburg, Germany

Spiral sampling patterns are well suited for fast MR applications like cardiac imaging, yet additional undersampling is necessary to achieve real-time temporal resolution. To still transform these data into images of high quality, a Variational Network is presented. The proposed model uses U-Nets for regularizing a gradient descent scheme, to reconstruct dynamic image series comparable to a segmented fully sampled cine investigation. The acquisition time for an entire stack from base to apex was below one minute; the overall reconstruction time was about 6 minutes.

Yifan Qi¹, Fusheng Wang¹, J. Jane Cao², and Yulee Li^2
1Computer Science, Stony Brook University, Stony Brook, NY, United States, ²St. Francis Hospital, DeMatteis Center for Cardiac Research and Education, Greenville, NY, United States

The presented work introduces a deep-learning cardiac MRI approach to quantitative assessment of ventricular volumes from raw MRI data without image reconstruction. As the information required for volumetric measurements is less than that for image reconstruction, ventricular function may be assessed with less MRI data than conventional image-based methods. This offers the potential to improve temporal resolution for quantitatively imaging cardiac function.

Qing Zou¹, Abdul Haseeb Ahmed¹, Prashant Nagpal¹, Rolf Schulte², and Mathews Jacob^1
1University of Iowa, Iowa City, IA, United States, ²GE Global Research, Munich, Germany

Cardiac disease is a frequent comorbidity and cause of death in subjects with compromised pulmonary function (e.g. COPD). High-resolution and free breathing cine imaging sequences are necessary for the evaluation of cardiac function in such subjects. Several self-gated and manifold algorithms were introduced for free-breathing MRI with good success. The main focus of this work is to further reduce the scan time of current manifold methods. We introduce a novel unsupervised deep generative manifold framework to recover free-breathing cine data from around 7seconds/slice, which will facilitate the acquisition of the whole volume in under two minutes.

Junyu Wang¹, Ruixi Zhou¹, and Michael Salerno^1,2,3
1Biomedical Engineering, University of Virginia, Charlottesville, VA, United States, ²Medicine, University of Virginia, Charlottesville, VA, United States, ³Radiology, University of Virginia, Charlottesville, VA, United States

Cardiac real-time cine imaging is valuable for patients who cannot hold their breath or have irregular heart rhythms. Spiral acquisitions, which provide high acquisition efficiency, make high-resolution cardiac cine real-time imaging feasible. However, the reconstruction for under-sampled non-Cartesian real-time imaging is time-consuming, and hence cannot provide rapid feedback. We sought to develop a DEep learning-based rapid Spiral Image REconstruction technique (DESIRE) for spiral real-time cardiac cine imaging with free-breathing, to provide fast and high-quality image reconstruction and make rapid online reconstruction feasible. High image quality was demonstrated using the proposed technique for healthy volunteers and patients.

Huili Yang^1,2, Amanda Lynn DiCarlo^1,2, Daming Shen^1,2, Hassan Haji-Valizadeh^1,2, Michael Markl^1,2, and Daniel Kim^1,2
1Department of Biomedical Engineering, Northwestern University, Evanston, IL, United States, ²Department of Radiology, Northwestern University, Chicago, IL, United States

While compressed sensing is a proven method for highly accelerating cardiovascular MRI, its lengthy reconstruction time hinders clinical translation. Deep learning is a promising method to accelerate reconstruction processing. We propose a generative adversarial network (GAN) with optimal loss terms for rapid reconstruction of 28.8-fold accelerated real-time phase-contrast MRI. Our results show that GAN reconstructs images 613 times faster than compressed sensing without significant loss in peak and mean velocity measurements and image sharpness.

Ria Garg^1,2, Elizabeth Hillier^2,3,4, Andrew Coristine^2,3,5, and Matthias G. Friedrich^2,3
1Cardiovascular imaging, RI-McGill University Health Center, Montreal, QC, Canada, ²Faculty of Medicine, McGill University Health Center, Montreal, QC, Canada, ³RI-McGill University Health Center, Montreal, QC, Canada, ⁴Faculty of Medicine, University of Alberta, Edmonton, Alberta, AB, Canada, ⁵MR applications and workflow, GE Healthcare, Montreal, QC, Canada

Cardiovascular magnetic resonance cine imaging is utilised to give comprehensive information on ventricular function.  Arrhythmia obstructs acquisition of these images, where the use of faster acquisition protocols with deep learning reconstruction methods may aid in solving the problem. We evaluated cine images of three patients in atrial arrhythmia, acquired using standard method; fast, variable density spatiotemporal sampling acquisition (VD kt) of one(1rr) or three(3rr) heart beats; and deep learning reconstruction of the same (DL-1rr,DL-3rr). Our results showed that undersampled techniques combined with deep-learning algorithms result in image quality improvements with no significant difference in quantitative values between all acquisition techniques.

Martin Schilling¹, Sebastian Rosenzweig^1,2, Moritz Blumenthal¹, and Martin Uecker^1,2,3
1Institute for Diagnostic and Interventional Radiology, University Medical Center Göttingen, Göttingen, Germany, ²DZHK (German Centre for Cardiovascular Research), Göttingen, Germany, ³Campus Institute Data Science (CIDAS), University of Göttingen, Göttingen, Germany

Cardiac segmentation is essential for analyzing cardiac function. Manual labeling is relatively slow, so machine learning methods have been proposed to increase segmentation speed and precision. These methods typically rely on cine MR images and supervised learning. However, for real-time cardiac MRI, ground truth segmentations are difficult to obtain due to lower image quality compared to cine MRI. Here, we present a method to obtain ground truth segmentation for real-time images on the basis of self-gated MRI (SSA-FARY).

Ummul Afia Shammi¹, Zhijian Luan², Jia Xu³, Aws Hamid⁴, Joanne Cassani⁴, Talissa A. Altes⁴, Robert P. Thomen⁴, and Steven R Van Doren^5
1Biomedical, Biological & Chemical Engineering, University of Missouri, Columbia, MO, United States, ²Institute for Data Science and Informatics, University of Missouri, Columbia, MO, United States, ³Biochemistry, University of Missouri, Columbia, MO, United States, ⁴Radiology, University of Missouri, Columbia, MO, United States, ⁵Biochemistry, Institute for Data Science, University of Missouri, Columbia, MO, United States

Two new methods of automatic, retrospective suppression of breathing motion appear to be effective for real-time cardiac MR (CMR) scans of free-breathing healthy volunteers. This may obviate the need for multiple breath-holds during CMR exams, allowing for a quicker, more comfortable experience for cardiac patients. The first method demonstrated corrects smaller breathing motions and has the advantage of correcting all cardiac cycles. The second method demonstrated successfully compensates the respiratory excursions of largest amplitude, common in short axis scans, by extracting cardiac cycles at end-inspiration or end-expiration.

Jack Allen^1,2, George Mathew¹, Miriam Conway¹, Sophie Jenkins¹, David Firmin^1,2, Jennifer Keegan¹, Sonya V. Babu-Narayan¹, and Peter Gatehouse^1
1Cardiovascular Biomedical Research Unit, Royal Brompton and Harefield NHS Trust, London, United Kingdom, ²National Heart and Lung Institute, Imperial College London, London, United Kingdom

3D Late Gadolinium-Enhanced (LGE) imaging is used to assess scarring in patients with Atrial Fibrillation (AF). Acquiring image data during every cardiac cycle allows a reasonable total scan duration but exacerbates ghosting artefacts caused by variable heart rates, such as those of patients with AF. Dynamic Inversion Time (TI) methods improve image quality by modifying the TI for each cardiac cycle. We present the initial stage of a patient study to validate a recently-proposed blood-focused dynamic TI algorithm. No improvement was found in comparison to the original algorithm. Future comparisons will include more patients with high R-wave interval variability.

Adam E. Goldman-Yassen¹, Maria J. Borja², Anna Derman², Eytan Raz², Duan Chen², Siddhant Dogra², Kai Tobias Block², and Seena Dehkharghani^2,3
1Department of Radiology, Children's Healthcare of Atlanta, Atlanta, GA, United States, ²Department of Radiology, New York University Langone Medical Center, New York, NY, United States, ³Department of Neurology, New York University Langone Medical Center, New York, NY, United States

GRASP is a fast and flexible MRI technique combining compressed sensing, parallel imaging, and golden-angle radial sampling. GRASP MRI obtained as part of routine clinical care without specific angiographic parameters can be displayed as 4D MRA images with high spatial and temporal resolution. It exhibits excellent performance and high reader confidence in dynamic angiography, using a highly flexible plug-and-play reconstruction and visualization pipeline. Further exploration of diagnostic accuracy in disease-specific applications is warranted.

KyungPyo Hong¹, Amanda L DiCarlo¹, Aggelos K Katsaggelos^1,2,3, Florian A Schiffers³, Cynthia K Rigsby^1,4, Hassan Haji-Valizadeh⁵, and Daniel Kim^1,6
1Radiology, Northwestern University Feinberg School of Medicine, Chicago, IL, United States, ²Electrical and Computer Engineering, Northwestern University, Evanston, IL, United States, ³Computer Science, Northwestern University, Evanston, IL, United States, ⁴Medical Imaging, Ann & Robert H. Lurie Children's Hospital of Chicago, Chicago, IL, United States, ⁵Internal Medicine, Cardiovascular Division, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA, United States, ⁶Biomedical Engineering, McCormick School of Engineering, Northwestern University, Evanston, IL, United States

While conventional density compensation function(DCF) performs sufficiently well for filtered backprojection(FBP) and radial k-space MRI when the Nyquist sampling condition is met and/or evenly-spaced view angles are used, it may perform poorly when sub-sampling and/or irrational-view angles are used. We propose an optimized DCF for the aforementioned conditions by calculating the density weights based on geometric properties of radial k-space sampling in a discrete environment, regardless of scan conditions such as data sizes and view angles.  Compared with standard DCF, the optimized DCF produces higher signal-to-noise ratio(SNR) in FBP (phantom) and more accurate flow metrics in 48-fold accelerated, phase-contrast MRI.  

V Emre Arpinar¹, L Tugan Muftuler¹, Andrew Nencka¹, Kevin Koch¹, and El-Sayed H Ibrahim^1
1Medical College of Wisconsin, Milwaukee, WI, United States

Myocardial strain imaging by MRI showed to be an early marker of changes in regional cardiac function before global cardiac function is affected. However, conventional MRI tagging techniques has the limitations of limited spatial resolution and complicated post-processing. Strain-encoding (SENC) showed to be a valuable technique for quantifying myocardial strain with high spatial resolution and simple post-processing. In this study, we provide preliminary results of echo-planar imaging (EPI)-based fast SENC imaging along with imaging artifacts and approaches to minimize their effects. The results showed the feasibility of EPI-based SENC with improved image quality and ultrafast acquisition.

Seong-Eun Kim¹, J Scott McNally¹, John A Roberts¹, Matthew Alexander ¹, Hediyeh Baradaran ¹, Gerald S Treiman², and Dennis L Parker^1
1UCAIR, Department of Radiology and Imaging Sciences, University of Utah, Salt Lake City, UT, United States, ²Department of Veterans Affairs, VASLCHCS, Salt Lake City, UT, United States

Dynamic contrast‐enhanced (DCE) MRI has been used for quantitative assessment of the neovascular architecture and perfusion properties in the carotid artery wall5. While DCE-MRI has great potential in plaque component characterization, it is presently limited by SNR, contrast, available resolution, and motion artifacts for carotid plaque application. In this work, we propose a new dynamic MRI method capable of high spatial and temporal resolution, and dynamic T1 mapping-based quantification of contrast agent concentration for the assessment of carotid intraplaque hemorrhage to overcome the barriers that prevent DCE-MRI from being the clinical standard for carotid IPH evaluation.

Joon Sik Park¹, Seung-Hong Choi², Chul-Ho Sohn², and Jaeseok Park^1,3
1Biomedical Engineering, Sungkyunkwan University, Gyunggido, Korea, Republic of, ²Seoul National University Hospital, Seoul, Korea, Republic of, ³Intelligent Precision Healthcare Convergence, Sungkyunkwan University, Gyunggido, Korea, Republic of

In this work, we propose a novel, joint functional segmentation and reconstruction directly from largely incomplete DCE measurements exploiting multi-scale vascular  heterogeneity priors to achieve high-definition whole brain DCE MRI. Experimental studies were performed in patients with brain tumor to investigate the feasibility of the proposed method in jointly visualizing vascular structures and functions from a single measurement after administering contrast agents. We demonstrated that the proposed method outperforms conventional methods in delineating angiographic structures, functional segment maps, and corresponding perfusion-permeability maps

Huiming Dong^1,2, Brian Raterman¹, Mariah Eisner³, Guy Brock³, Richard D White¹, Jean Starr⁴, Mounir Haurani⁴, Michael Go⁴, Patrick Vaccaro⁴, and Arunark Kolipaka^1,2
1Radiology, The Ohio State University Wexner Medical Center, Columbus, OH, United States, ²Biomedical Engineering, The Ohio State University, Columbus, OH, United States, ³Biomedical Informatics, The Ohio State University Wexner Medical Center, Columbus, OH, United States, ⁴Surgery, The Ohio State University Wexner Medical Center, Columbus, OH, United States

Abdominal aortic aneurysm (AAA) can result in life-threatening rupture. AAA diameter remains the only clinically useful parameter to predict growth and rupture risk. However, some high-risk AAAs can be overlooked due to their small diameters. AAA stiffness is associated with its extracellular matrix remodeling and thus potentially provides more relevant rupture prediction. Therefore, this study aims to estimate AAA stiffness using aortic MRE in patients with and without aneurysmal events. This is the first study to demonstrate that AAA stiffness and AAA/remote normal (AAA/RN) stiffness ratio are significantly lower in patients with aneurysmal events.

Olivier Jaubert^1,2, Grzegorz Kowalik², Javier Montalt-Tordera², Simon Arridge¹, Jennifer Steeden², and Vivek Muthurangu^2
1Department of Computer Science, UCL, London, United Kingdom, ²Centre for Cardiovascular Imaging, UCL, London, United Kingdom

Real-time spiral phase contrast MR is practical for free-breathing assessment of flow. However, for sufficient spatial and temporal resolutions it requires high acceleration rates leading to long reconstruction times. Here we propose to train a 3D U-Net with complex convolutions to accurately reconstruct phase data and flow curves from highly undersampled data. Prospectively acquired in-vivo data were reconstructed with similar image quality but ~4.6x faster than compressed sensing reconstructions which could improve workflow.

Eline Huizing¹, Richte C.L. Schuurmann², Frank F.J. Simonis³, Çağdaş Ünlü¹, Henri G.D. Leuvenink⁴, and Jean-Paul P.M. de Vries^2
1Department of Surgery, Northwest Clinics, Alkmaar, Netherlands, ²Department of Surgery, University Medical Center Groningen, Groningen, Netherlands, ³University of Twente, Enschede, Netherlands, ⁴Surgical Research Laboratory, University Medical Center Groningen, Groningen, Netherlands

Optimal treatment parameters of balloon angioplasty for revascularization such as inflation time and pressure are subject to debate. Limited data exists because accurate assessment of vascular function is challenging and therefore the influence procedure parameters have on arterial wall damage is largely unknown. This study shows that portable MRI at 0.125x0.125x2 mm³ resolution can be used for assessment of diameter and wall thickness during balloon angioplasty in ex vivo porcine arteries. Even after deflation and removal of the balloon the arterial diameter remained enlarged; this was coupled to a decrease in wall thickness.

Marylène Delcey^1,2,3,4, Pierre Bour^1,3,5, Isabelle Saniour⁶, Dounia El Hamrani^1,3,5, Valery Ozenne^1,3,5, Marie Poirier-Quinot⁶, and Bruno Quesson^1,3,5
1IHU Liryc, Electrophysiology and Heart Modeling Institute, Fondation Bordeaux Université, Pessac, France, ²Univ. Bordeaux, Centre de recherche Cardio-Thoracique de Bordeaux, U1045, Pessac, France, ³INSERM, Centre de recherche Cardio-Thoracique de Bordeaux, U1045, Bordeaux, France, ⁴Siemens Healthcare SAS, Saint-Denis, France, ⁵Univ. Bordeaux, Centre de recherche Cardio-Thoracique de Bordeaux, U1045, Bordeaux, France, ⁶IR4M, UMR8081, Université Paris-Sud/CNRS, Université Paris-Saclay, Orsay, France

In patients presenting cardiac electrical dysfunction, sub-millimetric resolution of CMR would help improving characterisation of the arrhytmogenic substrate. Here, a local coil was placed directly near the anatomic region to image to overpass the lack of sensitivity and selectivity of current conventionnal coils and reduce the voxel size. We implemented a motion compensation method that exploits the signal of micro-coils embodied on a catheter to retrospectively sort radial k-space data prior to ESPIRIT image reconstruction. Using this approach, we present high resolution (300µm), motion-resolved, image of the heart obtained in vivo in pig.

Ye Tian¹, Nam G. Lee², and Krishna S. Nayak^1
1Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, United States, ²Biomedical Engineering, University of Southern California, Los Angeles, CA, United States

We present a realistic cardiac MRI simulation framework, including concomitant effects, off-resonance, realistic coil-array and noise level. The general framework is flexible to different field strengths, sequences, and different motion patterns and ischemia models can be introduced. We demonstrate three usages of the framework, in comparing artifact/noise level at different field strengths, in optimization of real-time 3D trajectories and reconstruction, and capturing myocardial perfusion deficits. Real-time 3D cine and first-pass perfusion with high-resolution whole-heart coverage were tested feasible on a high-performance 0.55T system by simulation.

Pan Liu¹, Sidy Fall², and Olivier Baledent^2,3
1University of Picardie Jules Verne/CHIMERE EA 7516, Amiens, France, ²Facing Faces Institute/CHIMERE EA 7516, Amiens, France, ³Medical image processing department, University hospital Amiens, Amiens, France

Compared with CINE phase contrast MRI (CINE-PC), echo-planar imaging phase contrast (EPI-PC) can achieve real-time quantification of blood flow, with lower SNR. In this study, the pulsating model of the simulated cerebral vasculature was used to verify the accuracy of EPI-PC. The imaging time of EPI-PC was 62ms/image at 1.2*1.2mm² spatial resolution and FOV=100*60mm². The reconstructed EPI-PC flow curve was extracted by homemade post-processing software. After comparison with CINE-PC flow curve, it was concluded that EPI-PC can provide an average flow with less than 3% error, and its flow curve will be similar to the CINE-PC flow curve in shape.