Yue Pan^1,2, Juliet Varghese¹, Ning Jin³, Carmel Hayes⁴, Peter Speier⁴, Rizwan Ahmad^2,5, and Orlando Simonetti^1,2,6
1Davis Heart and Lung Research Institute, The Ohio State University Wexner Medical Center, Columbus, OH, United States, ²Department of Biomedical Engineering, The Ohio State University, Columbus, OH, United States, ³Siemens Medical Solutions USA, Malvern, PA, United States, ⁴Siemens Healthcare GmbH, Erlangen, Germany, ⁵Department of Electrical and Computer Engineering, The Ohio State University, Columbus, OH, United States, ⁶Department of Internal Medicine, The Ohio State University Wexner Medical Center, Columbus, OH, United States

In this study, the performance of Beat Sensor Cardiac (BSC) triggering was qualitatively and quantitatively compared with electrocardiogram (ECG) triggering. MR images typically acquired in a comprehensive exam including localizer, morphology, cine, 2D flow, parametric mapping and late gadolinium enhancement (LGE) were acquired using both BSC and ECG triggering in volunteers and patients. The overall image quality for BSC was equivalent to ECG. Quantitative measurements of function, flow, and parametric maps in healthy volunteers showed no significant differences except for peak aortic velocity.

Shuang Li¹, Wenjing Yang¹, Di Zhou¹, Yining Wang¹, Xiaohan Fan¹, Arlene Sirajuddin², Andrew E. Arai², Shihua Zhao¹, and Minjie Lu^1
1fuwai hospital, Beijing, China, ²National Heart, Lung and Blood Institute (NHLBI), National Institutes of Health (NIH), Bethesda, MD, United States

This study aims to identify the risk factors for adverse events in dilated cardiomyopathy (DCM) patients with LVEF≥35% and establish a scoring model to predict adverse event risk. 269 consecutive DCM patients with LVEF ≥35% who underwent gadolinium-enhanced cardiac magnetic resonance (CMR) imaging were enrolled in this study. The primary endpoint was a composite of SCD or aborted SCD. Secondary endpoints were all-cause mortality, heart transplantation and hospitalization for heart failure.  The nomogram we created using these parameters provides a novel way to clinically assess and risk stratify this patient population.

Yuchi Liu¹, Jesse Hamilton^1,2, Yun Jiang^1,2, and Nicole Seiberlich^1,2
1Department of Radiology, University of Michigan, Ann Arbor, MI, United States, ²Department of Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States

This work aims to extend rosette cardiac MRF to enable simultaneous myocardial T₁, T₂, and PDFF mapping from a single scan. The accuracy of rosette cMRF in T₁, T₂, and PDFF quantification is validated in the ISMRM/NIST system phantom and an in-house built fat fraction phantom, respectively. Studies in healthy subjects show that rosette cMRF potentially enables more accurate T₁ and T₂ measurements than spiral cMRF in the myocardium due to reduced fat contamination.

Yanan Zhao¹, Jianing Cui¹, Xueqian Liu², Tao Li¹, and Xiuzheng Yue^3
1Department of Radiology, Chinese People’s Liberation Army General Hospital, Beijing, China, ²Qinhuangdao Workers' Hospital, Qinhuangdao, China, ³Philips Healthcare, Beijing, China

Effective risk assessment and stratification are essential for the clinical management of ST-segment elevation myocardial infarction (STEMI) patients. CMR imaging has become a beneficial imaging modality to assess myocardial morphology, function and infarct characteristics simultaneously. This study quantitatively evaluated relationship between microvascular obstruction and regional myocardial strain by Cardiac MRI after ST-segment-elevation myocardial infarction. Our data showed that regional strain parameters may be s new noninvasive imaging markers allowing comprehensive evaluation regional myocardium deformation.

Pablo Villar-Calle¹, Jiahao Li², Alice Saffioti¹, Justin Johannesen¹, Katherine Tak¹, Rachel Meier¹, Jiwon Kim¹, Quynh A Truong³, Yi Wang², Pascal Spincemaille³, and Jonathan W Weinsaft^1
1Medicine, Weill Cornell Medicine, New York NY, United States, New York, NY, United States, ²Meinig School of Biomedical Engineering, Cornell University, Ithaca NY, United States, New York, NY, United States, ³Radiology, Weill Cornell Medicine, New York NY, United States, New York, NY, United States

This study tested cardiac MRI cardiac quantitative susceptibility mapping (QSM) for measurement of mitral annular calcification (MAC) using the reference of computed tomography (CT). QSM entailed a free-breathing ECG-triggered multi-echo 3D GRE sequence with diaphragmatic navigation via ECG-gating (1D respiratory navigator, 4mm window, acquisition time ~7 minutes, navigator efficiency 36%, spatial resolution 1.5×1.5×5mm3). Among 24 patients undergoing MRI and CT (mean 3.9±3.4 months), annular susceptibility decreased stepwise (p<0.001) and R2* increased (p<0.05) among patients stratified by CT mitral annular calcium score. Both QSM (AUC=0.950, p=0.005) and R2* (AUC=0.938, p=0.007) yielded excellent diagnostic performance for advanced mitral annular calcification on CT.

Manuel A. León¹, Rodrigo Salas¹, Sergio Uribe^2,3,4,5, and Julio Sotelo^1,3,5
1School of Biomedical Engineering, Universidad de Valparaíso, Valparaíso, Chile, ²Department of Radiology, School of Medicine, Pontificia Universidad Católica de Chile, Santiago, Chile, ³Millennium Nucleus in Cardiovascular Magnetic Resonance, CardioMR, Santiago, Chile, ⁴Institute for Biological and Medical Engineering, Schools of Engineering, Medicine and Biological Sciences, Pontificia Universidad Católica de Chile, Santiago, Chile, ⁵Biomedical Imaging Center, Pontificia Universidad Católica de Chile, Santiago, Chile

Cardiac contour delineation currently presents several proposals for automatic segmentation in short-axis MRI. However, most of them have drawbacks such as low accuracy with images from other MRI scanners, lack of description of the algorithms, and difficult access for clinical or research use. For this reason, we present VENTSEG, a deep learning framework for cardiac segmentation implemented in Matlab and Python. Our dice coefficient results segmenting images from the CNN validation set are 85.55% and 96.17% on anonymized clinical data. These results demonstrate the generalization of the framework on multidomain images.

Yuan Zheng¹, Lele Zhao², Zhongqi Zhang¹, Junpu Hu², Yu Ding¹, and Jian Xu^1
1UIH America, Inc, Houston, TX, United States, ²Shanghai United Imaging Healthcare Co. Ltd, Shanghai, China

GRE sequences benefit from increased SNR at high field and do not suffer from banding artifacts / SAR issues as bSSFP sequences. We have implemented an autocalibrated multiband retrospective GRE Cine sequence at 5 T. The autocalibrated multiband acquisition increases the efficiency of the sequence by reducing the scan time and simplifying the multiband imaging workflow. A compressed sensing reconstruction with implicit phase interpolation and temporal TV periodic boundary condition were implemented to further improve image quality. We have evaluated this application on volunteers and achieved high quality GRE Cine images.

Graeme Harris¹, Stephen Jermy^1,2, and Ernesta Meintjes^1,2
1Department of Human Biology, Division of Biomedical Engineering, University of Cape Town, Cape Town, South Africa, ²Cape Universities Body Imaging Center, University of Cape Town, Cape Town, South Africa

Respiratory motion of the heart is a fundamental challenge to cardiac MR imaging (CMR), frequently compensated for with breath-holding and acceptance-window methods. These methods are not always viable and result in inefficient acquisitions, creating longer scan times¹. Here, an adaptive Kalman-filter with a control system was implemented in a FLASH sequence to update the position of the slice during the imaging segment based on repeated navigator measurements acquired during the non-imaging segment. This predictive tracking results in a free-breathing sequence which is less strenuous for subjects, more efficient and reduces scan times.

Mehdi Ghafarinatanzi^1
1Polytechnique Montréal, Montreal, QC, Canada

The objective of this study is to investigate the myocardial 3D strain stress field of the patient-specific left ventricles in children with acute lymphoblastic leukemia (ALL) after treatment. The virtual field-finite element approach is applied to the 3D mesh generated tetrahedral for each patient-specific geometry over the cardiac cycle. The results illustrate the difference in distribution stress and displacement field for each LV subject.

Gaspar Delso¹, Pablo García-Polo², Margarita Gorodezky³, Ben Ariff⁴, Vicente Martínez de Vega⁵, and Javier Urmeneta^6
1GE Healthcare, Barcelona, Spain, ²GE Healthcare, Madrid, Spain, ³GE Healthcare, London, United Kingdom, ⁴Imperial College Healthcare NHS Trust, London, United Kingdom, ⁵Servicio de diagnóstico por la imagen, Hospital Universitario Quirónsalud, Madrid, Spain, ⁶Servicio de cardiología, Hospital Universitario Quirónsalud, Madrid, Spain

In this study, we explore the possibility of leveraging Deep Learning regularized reconstruction to enable lower flip angle MOLLI acquisition for myocardial T1 mapping. It has been shown in the past that lowering flip angle helps reduce various artifact sources, at the cost of lower signal to noise ratio. Regularized reconstruction can effectively manage image noise as well as increase feature sharpness. This hypothesis has been tested on a group of clinical patients referred for a cardiac MR exam.

Alper Ozan Turgut¹, Matthew Van Houten², Junyu Wang², Xue Feng², and Michael Salerno^3
1School of Medicine, University of Virginia, Charlottesville, VA, United States, ²Biomedical Engineering, University of Virginia, Charlottesville, VA, United States, ³Department of Medicine and Radiology, Stanford University, Stanford, CA, United States

Contrast-enhanced cardiac magnetic resonance (CMR) stress perfusion imaging shows excellent utility in evaluating coronary artery disease¹. Registering perfusion CMR image series is difficult due to the varying image contrast. Neural style transfer is a deep learning method used to transfer the “style” of one domain to another while preserving the content. Two neural style transfer networks were implemented in Python using TensorFlow and PyTorch. Training of each network was done using three, slice matched patient profiles and cine-like perfusion images were generated and registered. This method is compared to a KL-transform based registration approach.

Anthony Maroun¹, Justin Baraboo¹, Suvai Gunasekaran¹, Julia Hwang¹, Sophia Liu¹, Daniel Kim¹, Philip Greenland², Rod Passman³, Bradley Allen¹, Michael Markl¹, and Maurice Pradella^1,4
1Department of Radiology, Northwestern University Feinberg School of Medicine, Chicago, IL, United States, ²Department of Preventive Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL, United States, ³Department of Cardiology, Northwestern University Feinberg School of Medicine, Chicago, IL, United States, ⁴Department of Radiology, University Hospital Basel and University of Basel, Basel, Basel, Switzerland

The biplane area-length method is routinely used in clinical settings for left atrial volume (LAV) estimation. This technique is time-resolved but relies on geometrical assumptions of an ellipsoidal shape to estimate the LAV. In contrast, quantification derived from 3D segmentations on late gadolinium-enhanced magnetic resonance imaging and contrast-enhanced magnetic resonance angiography are static but do not rely on geometric assumptions. We compared the LAV estimation from these 3 methods in patients with atrial fibrillation and found a significant underestimation by the biplane area-length technique, indicating that this method may not capture the entire LAV in patients with complex anatomy.

Rui Guo¹, Zhensen Chen², Amine Amyar¹, Hossam El-Reiwady¹, Patrick Pierce¹, Beth Goddu¹, and Reza Nezafat^1
1Department of Medicine (Cardiovascular Division), Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA, United States, ²Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University, Shanghai, China

MOLLI T₁ mapping is routinely used for myocardial tissue characterization. Despite excellent precision, MOLLI has great T₁ underestimation and a long breath-holding time. We previously developed a deep learning-based approach (MyoMapNet) for myocardial T₁ mapping using four T₁-weighted images. MOLLI T₁ was used to train a fully connected neural network (FC), which resulted in a similar accuracy error as MOLLI. In this study, numerically simulated and phantom signals with the presence of different B0, flip angles, and heart rates were used to improve MyoMapNet T₁ accuracy. Evaluation showed that MyoMapNet T₁ could be improved and had higher precision than SASHA.