Verónica Aramendía-Vidaurreta¹, Pedro Macías-Gordaliza^2,3, Marta Vidorreta⁴, Rebeca Echeverria-Chasco¹, Gorka Bastarrika¹, Arrate Muñoz-Barrutia^2,3, and María Fernández-Seara^1
1Radiology, Clínica Universidad de Navarra, Pamplona, Spain, ²Universidad Carlos III de Madrid, Madrid, Spain, ³Instituto de Investigación Sanitaria Gregorio Marañón, Madrid, Spain, ⁴Siemens Healthineers, Madrid, Spain

Arterial spin labeling enables non-invasive quantification of myocardial perfusion. However, there is a need to improve its reproducibility by reducing subtraction errors due to motion and thus, avoiding signal blurring in the acquired low-resolution images. Here, we demonstrate the feasibility of quantifying MBF employing synchronized breathing techniques with FAIR-ASL in healthy subjects. Visually, motion is significantly reduced, MBF values are consistent with those reported in the literature and coefficient of variation for intrasession reproducibility is 13%.

Sina Amirrajab¹, Yasmina Al Khalil¹, Cristian Lorenz², Juergen Weese², and Marcel Breeuwer^1,3
1Biomedical Engineering Department, Eindhoven University of Technology, Eindhoven, Netherlands, ²Philips Research Laboratories, Hamburg, Germany, ³MR R&D - Clinical Science, Philips Healthcare, Best, Netherlands

This study investigates an approach to generate a realistic, heterogeneous database of simulated cardiac MR images to aid the development of fully automated and generalizable deep learning based segmentation algorithms, less sensitive to variability in CMR image appearance. XCAT phantoms were used to create the virtual population by altering the heart position and geometry and MRXCAT approach was improved to simulate more organs. Images simulated in this study were quantitatively and qualitatively comparable to real CMR images acquired by two different sites and vendors. Initial experiments using such a heterogeneous image dataset show a positive impact on the segmentation performance.

Seung Su Yoon^1,2, Elisabeth Hoppe¹, Michaela Schmidt², Christoph Forman², Teodora Chitiboi³, Puneet Sharma³, Christoph Tillmanns⁴, Andreas Maier¹, and Jens Wetzl^2
1Department of Computer Science, Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany, ²Magnetic Resonance, Siemens Healthcare GmbH, Erlangen, Germany, ³Siemens Medical Solutions USA, Inc, Princeton, NJ, United States, ⁴Diagnostikum Berlin, Berlin, Germany

The detection of a window with the least motion, e.g. end-systolic or end-diastolic resting phases (RPs) within the cardiac cycle is necessary for data acquisition for static cardiac imaging. In the current workflow, it is manually performed on CINE images by visual inspection. To automate the workflow, we propose an improved Deep-Learning-based automated localized RP detection. While the first step is responsible to localize the anatomy of interest, the second is to quantify motion within the target, and third is to classify RPs. We validated the system with a prospective volunteer study and achieved accuracy in the range of 28ms.

Seung Su Yoon^1,2, Michaela Schmidt², Bernd J Wintersperger^3,4, Teodora Chitiboi⁵, Puneet Sharma⁵, Christoph Tillmanns⁶, Andreas Maier¹, and Jens Wetzl^2
1Department of Computer Science, Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany, ²Magnetic Resonance, Siemens Healthcare GmbH, Erlangen, Germany, ³Department of Medical Imaging, University Health Network, Toronto, ON, Canada, ⁴Department of Medical Imaging, University of Toronto, Toronto, ON, Canada, ⁵Siemens Medical Solutions USA, Inc, Princeton, NJ, United States, ⁶Diagnostikum Berlin, Berlin, Germany

In Cardiac MRI, Late gadolinium enhancement (LGE) imaging is generally performed for the assessment of myocardial viability. As LGE is based on inversion recovery techniques, the correct myocardial nulling is necessary for image contrast optimization. In current clinical practice, it is done by visual evaluation. As it required user expertise and interaction, an automated inversion time selection is proposed. The Deep-Learnig-based system to detect the null point of inversion time was successfully demonstrated in all datasets comparing with two expert annotations.

Peng Liu¹ and Lijia Wang^1
1University of Shanghai for Science and Technology, Shanghai, China

It is essential to segment right ventricle (RV) for evaluating cardiac functional parameters of cardiac diseases in clinical diagnosis and prognosis. However, the complex structure of RV makes traditional segmentation methods not so effective in right ventricular segmentation. A new Dense and Multi-scale U-net deep learning method is proposed to segment right ventricle in cine cardiac magnetic resonance (CMR) short-axis images automatically, which shows high coincidence and small difference with manual segmentation and is promising for diagnosis and analysis of clinical cardiac diseases.
 

XIAODAN ZHAO¹, LIWEI HU², RONG ZHEN OUYANG², RU SAN TAN^1,3, PING CHAI⁴, MARIELLE FORTIER^3,5, SHUO ZHANG⁶, WEN RUAN¹, SHUANG LENG¹, JUN-MEI ZHANG^1,3, BRYANT JENNIFER¹, LYNETTE LS TEO⁴, ROB VAN DER GEEST⁷, TENG HONG TAN^3,5, JAMES YIP⁴, JU LE TAN^1,3, YUMIN ZHONG², and LIANG ZHONG^1,3
1National Heart Centre Singapore, Singapore, Singapore, ²Shanghai Children’s Medical Centre, Shanghai Jiaotong University School of Medicine, ShangHai, China, ³Duke-NUS Medical School, Singapore, Singapore, ⁴National University Hospital Singapore, Singapore, Singapore, ⁵KK Women’s and Children’s Hospital, Singapore, Singapore, ⁶Philips Germany, Humburg, Germany, ⁷Department of Radiology, Leiden University Medical Center, Leiden, Netherlands

Whole-heart 4D cardiovascular magnetic resonance (CMR) phase-contrast flow measurement enables qualitative and quantitative assessment of intra-cardiac flow. Its feasibility was investigated in 12 pediatric and 13 adult patients with repaired Tetralogy of Fallot (rTOF). Left ventricular (LV) kinetic energy (KE) and pathline-derived flow components were analyzed. KE parameters were significantly correlated with age, but only LV systolic KE remained significantly different after indexing to LV end-diastolic volume. Pediatric rTOF patients had similar LV ejection fraction, indexed LV volumes and mass, but significantly reduced indexed LV systolic KE, retained inflow and increased delayed ejection flow compared with adults.

jing zhu¹, lei wang², and fabao gao^2
1sichuan university, chengdu city, China, ²sichuan university, chengdu, China

 To assess the diagnostic value of CMR Tissue-Tracking(TT) for strain analysis in a rat model of acute myocarditis and the strain’s association with myocardial impairment. Experimental autoimmune myocarditis(EAM) was induced in 16 male rats, 10 rats served as control. Rats were scanned at 7T MRI 21 days after model induction, using cine-FLASH sequence and late gadolinium enhancement imaging. Myocardial strain values significantly reduced in EAM rats, compared with the controls. The area under the curve(AUC) of myocardial strain was excellent. Quantitative analysis of late gadolinium enhancement(LGE) showed a significant correlation with myocardial strain.

Laura Boehmert¹, Helmar Waiczies², Andre Kuehne², Celal Oezerdem¹, Sonia Waiczies¹, Ludger Starke¹, Min-Chi Ku¹, Andreas Pohlmann¹, Paula Ramos Delgado¹, Erdmann Seeliger³, and Thoralf Niendorf^1,2,4,5
1Berlin Ultrahigh Field Facility (B.U.F.F.), Max Delbrück Center for Molecular Medicine in the Helmholtz Association, Berlin, Germany, ²MRI.TOOLS GmbH, Berlin, Germany, ³Institute of Vegetative Physiology, Charité University Medicine, Berlin, Germany, ⁴DZHK (German Centre for Cardiovascular Research), partner site Berlin, Berlin, Germany, ⁵Experimental and Clinical Research Center, a joint cooperation between the Charité Medical Faculty and the Max Delbrück Center for Molecular Medicine, Berlin, Germany

We propose probing of tissue sodium concentrations in the heart using ²³Na MRI for adding a very useful dimension to our understanding of cardiac disorders. Changes in the sodium concentration might be indicative of early pathophysiological changes in such diseases. This work focuses on the development, evaluation and application of a quadrature birdcage RF resonator tailored for cardiac ²³Na MRI at 9.4T with the goal to provide a uniform excitation profile.

Eric Kenneth Gibbons^1,2, Ye Tian³, Qi Huang⁴, Akshay Chaudhari⁵, and Edward DiBella^2,4
1Electrical and Computer Engineering, Weber State University, Ogden, UT, United States, ²Radiology and Imaging Sciences, University of Utah, Salt Lake City, UT, United States, ³Physics, University of Utah, Salt Lake City, UT, United States, ⁴Biomedical Engineering, University of Utah, Salt Lake City, UT, United States, ⁵Radiology, Stanford University, Stanford, CA, United States

Current acquisition strategies in cardiac perfusion MRI rely on non-uniform sampling that is highly undersampled in spatial and temporal domains.  While iterative reconstruction methods are able to reconstruct such data reasonably well, reconstruction speeds are prohibitively long. This abstract applies novel deep learning approaches to accelerate reconstruction speeds relative to iterative algorithms with comparable image quality.  Validation is performed through the calculation of a perfusion index.

Stefano Buoso¹, Christian T Stoeck¹, Johanna Stimm¹, and Sebastian Kozerke^1
1Institute for Biomedical Engineering, ETH Zurich, Zurich, Switzerland

We show the feasibility of embedding physiological coordinates and directions into a left ventricle anatomical shape model using Proper Orthogonal Decomposition. The volumetric anatomical mesh and the physiological parametrization can be personalized directly from the selection of control points on MR cardiac images. This approach provides a consistent way of augmenting low-resolution data using features from high-resolution datasets. Additionally, the physiological parametrization is automatically adapted to each specific case without any additional calculation steps. This simplifies the processing of clinical images and, particularly, strain calculations and microstructural analysis that require the definition of the physiological parametrization.

Yue Gao¹, Zhi-gang Yang¹, Rui Shi¹, and Xi Liu^1
1radiology, west china hospital of sichuan university, chengdu, China

Myocardium steatosis was positively associated with decreased myocardial deformation and perfusion dysfunction.

Jin Zhang¹, Xiaoyi Chen², Huilin Zhao¹, Bin Cui³, Xiao Li¹, Beibei Sun¹, Xiaosheng Liu¹, Zhongshuai Zhang⁴, Xihai Zhao⁵, and Jianrong Xu^1
1Renji Hospital, Shanghai Jiaotong University, Shanghai, China, ²Beijing Geriatric Hospital, Beijing, China, ³Aerospace Center Hospital, Beijing, China, ⁴Siemens Healthcare, Shanghai, China, ⁵Biomedical Engineering & Center for Biomedical Imaging Research, Tsinghua University, Beijing, China

This study employed isotropic high-resolution SPACE or VISTA sequence to explore the association between intracranial ICA steno-occlusive disease caused by atherosclerosis and diffuse wall thickening (DWT) in ipsilateral petrous ICA. The results indicated that the diffuse wall thickening of petrous ICA is commonly seen among patients with steno-occlusive disease causing by atherosclerotic plaque in intracranial ICA, especially in occluded intracranial ICA.

lian luo¹, shuai liu², Peirong Jiang³, Tan Gong¹, Yuqian Mei¹, Fei Shang¹, and Xihai Zhao^2
1Department of Biomedical Engineering, Beijing Institute of Technology, Beijing, China, ²Tsinghua University, Beijing, China, ³Fujian Medical University Union Hospital, Fujian, China

To evaluate carotid lumen segmentation results on bifurcation geometry quantification, bifurcation angle and mean centerline distance were calculated. The intra-class correlation coefficient (ICC), linear regression and Bland-Altman plot of bifurcation angle, as well as the mean centerline distance indicated that the segmentation results had strong correlation with golden standard. The segmentation results could be used on geometry quantification.

Yuanbo Yang¹, Zhonghao Li², Yihao Guo¹, Yingjie Mei³, Ming Zhao², Guoxi Xie⁴, and Yanqiu Feng^1
1Guangdong Provincial Key Laboratory of Medical Image Processing & Key Laboratory of Mental Health of the Ministry of Education, School of Biomedical Engineering, Southern Medical University, Guangzhou, China, ²Department of Pathophysiology, Key Lab for Shock and Microcirculation Research of Guangdong, School of Basic Medical Sciences, Southern Medical University, Guangzhou, 510515, China, Guangzhou, China, ³philips healthcare, Guangzhou, China, ⁴Department of Biomedical Engineering, School of Basic Sciences, Guangzhou Medical University, Guangzhou, China

  Carotid atherosclerosis is a degenerative disease of the arterial wall, which can result in predisposition to cerebral thrombo-embolic stroke1. The combination of DANTE and VF-RARE (DANTE-VF-RARE) can suppress flow signal effectively. It was used for vessel wall imaging of human in the previous studies4. However, the feasibility of DANTE-VF-RARE sequence remains unclear for imaging of carotid for apoE-/- mouse. This work aims to develop 3D DANTE-VF-RARE on 7Tesla (T) and investigate its feasibility for vessel wall imaging of apoE-/- mouse.  

Mariah RR Daal¹, Bram F Coolen¹, Dorita THPM Dekkers¹, David Hauteman², Rob CI Wüst³, and Gustav J Strijkers^1
1Department of Biomedical Engineering, Amsterdam University Medical Center, University of Amsterdam, Amsterdam, Netherlands, ²Medis medical imaging systems B.V., Leiden, Netherlands, ³Faculty of Behavioural and Movement Sciences, Vrije Universiteit, Amsterdam, Netherlands

The current classification of heart failure based on diastolic and systolic dysfunction limits the understanding of the time course of heart failure with preserved left ventricular ejection fraction (HFpEF), and new parameters are warranted to assess cardiac function. Here, we assessed global longitudinal strain (GLS), as well as hemodynamic forces (HDF) in a mouse model of diabetes (db/db) at two time points, with the aim to study the prognostic value of these new parameters for cardiac dysfunction.

Seong-Jae Park¹, Jong-Hyun Yoon², and Chang-Beom Ahn^1
1Kwangwoon University, Seoul, Republic of Korea, ²Gachon University, Incheon, Republic of Korea

Compressed-sensing cardiovascular CINE MRI was performed using deep artificial neural network and transfer learning. Transfer learning is a method to use weights obtained from previous learning as initial weights for current learning to improve generality and performance of the neural network. When learning data is limited, it is useful for generalization by using previous learning along with other data. It also reduces learning time by 80 to 98 percent. And to prevent modification of measurement data by Deep learning, K-space correction was added as a post-processing process.

Elisabeth Hoppe¹, Jens Wetzl², Seung Su Yoon¹, Manuel Schneider², Bernhard Stimpel¹, Alexander Preuhs¹, and Andreas Maier^1
1Department of Computer Science, Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany, ²Magnetic Resonance, Siemens Healthcare, Erlangen, Germany

For continuous cardiac CINE acquisitions, cardiac binning of the data is necessary, which is done either using ECG-gating or hand-crafted postprocessing methods. To overcome these limitations, we propose a deep learning classifier to detect R-waves from repeated 1-D superior-inferior projections of the imaged data. After training with R-wave positions from the ECG signal as ground-truth data, detection of R-waves is possible without additional ECG-gating or hand-crafted features and can be used for retrospective cardiac binning. Our first proof-of-concept achieves a high accuracy of over 91% on previously unseen cardiac CINE data.

Sona Ghadimi¹, Xue Feng¹, Craig H. Meyer¹, and Frederick H. Epstein^1
1Biomedical Engineering, University of Virginia, Charlottesville, VA, United States

DENSE myocardial strain imaging is a method wherein tissue displacement is encoded in the image phase. Myocardial segmentation and phase unwrapping are two key steps in quantitative displacement and strain analysis of DENSE images. Prior DENSE analysis methods for segmentation and phase unwrapping were semi-automated techniques, requiring user intervention. In this study, we developed deep learning (DL) methods for fully automated myocardial segmentation and phase unwrapping for short-axis DENSE images. Quantitative and qualitative evaluations show promising results for the proposed DL-based segmentation and phase unwrapping methods, eliminating all manual steps needed for fully automatic DENSE strain analysis.

Yan-Chi Ivy Chan^1,2, Jay B. Patel³, Wai-Yan Ryana Fok^1,2, Kwannapas Saengsin^1,4, Andrew J. Powell¹, and Mehdi H. Moghari^1
1Department of Cardiology, Boston Children's Hospital and the Department of Pediatrics, Harvard Medical School, Boston, MA, United States, ²Department of Informatics, Technical University of Munich, Garching, Germany, ³Massachusetts Institute of Technology, Cambridge, MA, United States, ⁴Department of Pediatrics, Chiang Mai University, Chiang Mai, Thailand

We have developed and evaluated a new real-time acquisition and reconstruction technique for 2D steady-state free precession cine cardiac magnetic resonance imaging (MRI). Patient study is presented to validate the proposed technique.

Davide Piccini^1,2,3, Aurélien Maillot^1,2, John Heerfordt^1,2, Dimitri Van De Ville^4,5, Juerg Schwitter⁶, Matthias Stuber^2,7, Jonas Richiardi^1,2, and Tobias Kober^1,2,3
1Advanced Clinical Imaging Technology, Siemens Healthcare, Lausanne, Switzerland, ²Department of Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, ³LTS5, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, ⁴Institute of Bioengineering/Center for Neuroprosthetics, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, ⁵Department of Radiology and Medical Informatics, University Hospital of Geneva (HUG), Geneva, Switzerland, ⁶Division of Cardiology and Cardiac MR Center, University Hospital of Lausanne (CHUV), Lausanne, Switzerland, ⁷Center for Biomedical Imaging (CIBM), Lausanne, Switzerland

Learning anatomical characteristics from large databases of radiological data could be leveraged to create realistic representations of a specific subject’s anatomy and to provide a personalized clinical assessment by comparison to the acquired data. Here, we extracted 2D patches containing the descending aorta from 297 3D whole-heart MRI acquisitions and trained a Wasserstein generative adversarial network with a gradient penalty term (WGAN-GP). We used the same network to generate realistic versions of the aortic region on masked real images using a loss function that combines a contextual and a perceptual term. Results were qualitatively assessed by an expert reader.

Markus J. Ankenbrand¹, David Lohr¹, Tobias Wech², and Laura M. Schreiber^1
1Chair of Cellular and Molecular Imaging, Comprehensive Heart Failure Center (CHFC), University Hospital Würzburg, Würzburg, Germany, ²Department of Diagnostic and Interventional Radiology, University Hospital Würzburg, Würzburg, Germany

Training of neural networks for segmentation of CMR images requires large amounts of labeled data and network generalization is biased by training data characteristics. We used a specialized network to label a heterogeneous, publicly available dataset of 1140 cine images with known left-ventricular volumes. We evaluated the performance of this network using true and predicted volumes and trained another neural network on subjects with high prediction accuracy using extensive data augmentation. The resulting network outperforms the original one on the full dataset, even on subgroups where the original network fails, indicating great generalization and thus suitability for transfer learning applications.

Michael Baran Scott¹, Haben Baran Berhane², Kevin Ryan Kalisz¹, Tugce Agirlar Trabzonlu¹, Jeesoo Lee¹, Marci Messina¹, Chris Malaisrie¹, Patrick McCarthy¹, James Carr¹, Alexander J Barker³, Ryan Avery¹, and Michael Markl^1
1Northwestern University, Chicago, IL, United States, ²Lurie Children's Hospital of Chicago, Chicago, IL, United States, ³University of Colorado, Anschutz Medical Campus, Aurora, CO, United States

A convolutional neural network (CNN) originally implemented for time-averaged 3D segmentation of the thoracic aorta from 4D flow MRI was retrained to generate time-resolved segmentations without generating additional reference data. To validate the segmentations, automatically generated time-resolved segmentations were compared against two 2D cine acquisitions in 20 patients. The CNN achieved average Dice scores 0.87±0.04 and 0.88±0.04 for candy-cane and cross-section views of the aorta across all patients and timepoints. Automated time-resolved segmentation of 4D flow MRI data will enable calculation of metrics such as wall shear stress and aortic compliance that are sensitive to wall location.

Thomas Joyce¹, Stefano Buoso¹, Yuerong Xu¹, and Sebastian Kozerke^1
1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland

There has been considerable success applying deep learning to cardiac MRI segmentation. However, segmentation masks have several shortcomings. In particular, they are discrete (voxel-based) representations of continuous anatomy, and are hence not suitable for use in biomechanical simulation. Mesh representations can potentially overcome both of these drawbacks. We demonstrate an approach to predicting left-ventricular (LV) meshes from short-axis MR images. The proposed approach: (i) works robustly on data with differing slice numbers, slice thicknesses, and left-ventricular coverage, (ii) has no test-time optimisation loop, but rather directly predicts the mesh from a mask, and, (iii) generalises to real data with pathology.

Yoon-Chul Kim¹ and Yeon Hyeon Choe^2
1Clinical Research Institute, Samsung Medical Center, Sungkyunkwan Univ. School of Medicine, Seoul, Korea, Republic of, ²Department of Radiology, Samsung Medical Center, Sungkyunkwan Univ. School of Medicine, Seoul, Korea, Republic of

Radiologic diagnosis of myocardial late gadolinium enhancement (LGE) is often represented as a description of lesion characteristics and distributions in the standard 16- (or 17-) segment myocardial model. In this study, we used short-axis LGE images and 16-segment labeled results from 66 patients with coronary artery disease and non-ischemic heart disease and trained a deep convolutional neural network (CNN) model. Short-axis images were transformed to polar coordinates after identification of the LV center point and anterior RV insertion point. The proposed method does not require manual delineation of the lesions and potentially enables automatic diagnosis of myocardial viability.

Botian Xu¹, Yaqiong Chai¹, and John C. Wood^1,2
1Department of Biomedical Engineering, University of Southern California, Los Angeles, CA, United States, ²Division of Cardiology, Children's Hospital Los Angeles, Los Angeles, CA, United States

Tagging in cardiac MRI could cause great challenges to subsequent downstream analysis. Modeling tag removal task as an image recovery problem, we propose a GAN-based method constrained by edge prior, termed EG-GAN, to remove tags progressively without model collapse, and it outperforms conventional approaches.

Ricardo A Gonzales^1,2, John Onofrey¹, Jérôme Lamy¹, Felicia Seemann^3,4, Einar Heiberg^3,4,5, and Dana C Peters^1
1Department of Radiology and Biomedical Imaging, Yale University, New Haven, CT, United States, ²Department of Electrical Engineering, Universidad de Ingenieria y Tecnologia, Lima, Peru, ³Department of Clinical Physiology, Lund University, Lund, Sweden, ⁴Department of Biomedical Engineering, Lund University, Lund, Sweden, ⁵Wallenberg Center for Molecular Medicine, Lund University, Lund, Sweden

Diastolic dysfunction is assessed by measurement of mitral annular (MA) early diastolic velocity (e’), commonly performed in echocardiography. Similar measurements can be obtained with valvular plane tracking in MRI long-axis cines. These measurements have been validated and have good reproducibility, yet manual MA points annotations are required. In this work we present a machine learning convolutional neural network with a residual architecture for automatic annotation of MA points in MRI long-axis cine images of the 2 and 4-chamber views. The landmark tracking allowed a fast and accurate evaluation of diastolic parameters improving the clinical applicability of MRI for diastolic assessment.

Julius Frederik Heidenreich¹, Tobias Gassenmaier¹, Markus Ankenbrand², David Lohr², Thorsten Alexander Bley¹, and Tobias Wech^1
1Department of Diagnostic and Interventional Radiology, University Hospital Würzburg, Würzburg, Germany, ²Congestive Heart Failure Center, University Hospital Würzburg, Würzburg, Germany

Quantitative analysis of scar tissue in late gadolinium enhancement (LGE) cardiac magnetic resonance imaging (CMR) typically requires manual or at best semi-automatic segmentation by a trained physician. To supersede this time-consuming and tedious task, a convolutional neural network with a U-Net architecture and a ResNet34 backbone was trained for semantic segmentation of scar tissue in LGE CMR. The predictions of the proposed model yielded high performance for the detection of focal scar tissue and bears thus potential for fully automated and consequently time-efficient post-processing.  

Qiaoyi Xue¹, Jianmin Yuan¹, Haodong Qin², Hui Liu¹, Ran Huo³, and Huishu Yuan^3
1Central Research Institute, United Imaging Healthcare Group, United Imaging Healthcare, Shanghai, China, ²Magnetic Resonance Business Unit, United Imaging Healthcare, Shanghai, China, ³Department of Radiology, Peking University Third Hospital, Beijing, China

Time-of-flight (TOF) sequence is a non-contrast enhancing blood vessel imaging technique critical for vessel wall analysis. However, the periodical pulsation in the vessels, especially arteries, often causes pulsatile artifact along the phase encoding direction in the image. This degrades image quality and brings difficulty for diagnosis. The purpose of this work is to develop a convolutional neural network (CNN) method to reduce the pulsatile artefact in TOF sequence.

Cian Michael Scannell¹, Adriana Villa¹, Stefano Figliozzi¹, Jack Lee¹, Mikto Veta², Marcel Breeuwer^2,3, and Amedeo Chiribiri^1
1King's College London, London, United Kingdom, ²Eindhoven University of Technology, Eindhoven, Netherlands, ³Philips Healthcare, Best, Netherlands

Late gadolinium enhancement MRI is crucial tool for guiding the management of patients with known/suspected myocardial infarction. Currently, clinical practice relies on the visual inspection of these images but there has been extensive work on semi-automated approaches for quantifying scar. Their utility is however limited by the time-intensive manual interactions involved, including the drawing of the myocardial contours. In this work, deep learning methodology is employed to automatically achieve the previously manual steps and these segmentations are used as input to a completely unsupervised scar quantification algorithm. This allows automatic, fast and reproducible quantification of regions of scar.

Ruponti Nath¹, Sean Callahan¹, Narayana Singam², Marcus Stoddard², and Amir Amini^1
1ECE, University of Louisville, Louisville, KY, United States, ²Cardiovascular Medicine, University of Louisville, Louisville, KY, United States

We propose a framework for accelerated reconstruction of 2D phase contrast MRI from undersampled K space by using deep convolutional neural networks.  The reconstruction problem is considered as a de-aliasing problem in complex spatial domain. A U-net architecture  was trained and tested on 4D flow MRI data in 10 patients with aortic stenosis and 4 healthy volunteers. The reconstructed complex two channel image showed that the U-net is able to unaliase the undersampled flow images with resulting magnitude and phase difference images showing good agreement with the fully sampled magnitude and phase images.

Patrick Metze¹, Hao Li², Tobias Speidel², Ina Vernikouskaya¹, and Volker Rasche^1,2
1Department of Internal Medicine II, University Ulm Medical Center, Ulm, Germany, ²Core Facility Small Animal Imaging (CF-SANI), Ulm University, Ulm, Germany

Reconstruction methods incorporating Deep Learning have gained a lot of traction in the recent past. However, most of these methods have been evaluated in human MRI. In this work, we show the feasibility of deep-learning artifact removal for the tiny golden angle radial trajectory in rodent cardiac MRI and validate two approaches against a Compressed Sensing reconstruction and the gated reference standard. The deep-learning based methods achieve acceptable visual image quality and exhibit only slight, but for one method significant, differences in the functional analysis.

Jack JJJ Miller^1,2,3, Matthew Kerr², Ladislav Valkovic^3,4, Kerstin Timm², Justin Lau^2,3, Paul A. Bottomley⁵, Lisa Heather², and Damian Tyler^3
1Department of Physics, University of Oxford, Oxford, United Kingdom, ²Department of Physiology, Anatomy and Genetics, University of Oxford, Oxford, United Kingdom, ³OCMR, Cardiovascular Medicine, University of Oxford, Oxford, United Kingdom, ⁴Department of Imaging Methods, Institute of Measurement Science, Slovak Academy of Sciences, Bratislava, Slovakia, ⁵Radiology and Radiological Science, Johns Hopkins, Baltimore, MD, United States

Saturation Transfer 31P MRS enables the quantification of the rate of ATP synthesis and degradation. In this work, we use quasi-adiabatic optimal-control designed single-band (PCr) and dual-band (Gamma ATP + Pi) saturation schemes to separately estimate $$$k_f$$$ and $$$k_r$$$ in the control and diabetic perfused rat heart at 12T. This scheme is rapid, B1 insensitive, and suitable for use at ultrahigh field. We demonstrate profound metabolic alterations in the diabetic myocardium, with a reduction in $$$k_r$$$ and ATP concentration.

Galen Durant Reed¹, JaeMo Park², Junjie Ma², Crystal Harrison², Rolf Schulte³, Albert Chen ⁴, Salvador Pena², Jeannie Baxter², Kelley Derner², Maida Tai², Dean Sherry², Vlad G Zaha², and Craig R Malloy^2
1GE Healthcare, Dallas, TX, United States, ²University of Texas Southwestern Medical Center, Dallas, TX, United States, ³GE Healthcare, Munich, Germany, ⁴GE Healthcare, Toronto, ON, Canada

Initial human hyperpolarized ¹³C cardiac imaging experiments utilized a frequency-selective, single-shot spiral pulse sequence for encoding, and a rigid 8 channel receiver array paired with a Helmholtz clamshell transmitter. In phantom studies, we evaluate this RF hardware along with flexible transmit / receiver designs and test these in healthy volunteers. Furthermore, as pointed out previously, the detected circumferential distribution of [¹³C]bicarbonate around the myocardium is often non-uniform in pigs, thus confounding the translation of this technique for regional metabolism evaluation. We observed this artifact in human volunteers and explore its possible sources.

Sungtak Hong¹, Kyungpyo Hong², Austin E Culver¹, Bradley D Allen¹, Daniel C Lee³, and Daniel Kim^1
1Radiology, Northwestern University, Chicago, IL, United States, ²Translational and Molecular Imaging Institute and Department of Radiology, Icahn School of Medicine at Mount Sinai, New York, NY, United States, ³Division of Cardiology, Internal Medicine, Northwestern University, Chicago, IL, United States

CMR measurements in CIED patients are critical but challenging due to susceptibility artifacts. We described the development of highly accelerated, real-time, free-breathing cine pulse sequence using compressed sensing with variable density “lattice-like” k-space sampling scheme. Comprehensive comparison was done to estimate performance against clinical standard breath-hold cine pulse sequence.         

Yan Wen^1,2, Jonathan W. Weinsaft³, Thanh D. Nguyen², Jiwon Kim³, Yi Wang^4,5, and Pascal Spincemaille^2
1Meinig School of Biomedical Engineering, Cornell University, new york, NY, United States, ²Radiology, Weill Cornell Medicine, New York, NY, United States, ³Medicine, Weill Cornell Medicine, New York, NY, United States, ⁴Meinig School of Biomedical Engineering, Cornell University, New York, NY, United States, ⁵Radiology, Weill Cornell Medicine, new york, NY, United States

Previous studies have demonstrated noninvasive cardiac chamber blood oxygenation measurement using T2 and QSM. This work shows the initial comparison between the two approaches for such measurements in healthy volunteers. 

Xin Dong^1,2, Olumide Awelewa ^1,2, Graham Galloway², Viral Chikani³, Rob Robergs¹, and Arnold Ng^4,5
1Queensland University of Technology, Brisbane, Australia, ²Translational Research Institute, Woolloongabba, Australia, ³Endocrinology, Princess Alexandra Hospital, Woolloongabba, Australia, ⁴Cardiology, Princess Alexandra Hospital, Woolloongabba, Australia, ⁵The University of New South Wales, Sydney, Australia

VARPRO is a novel fat water separation sequence that shows promise for myocardial lipid characterization. Two questions were addressed in this study: 1. Does VARPRO provide the same myocardial fat-fraction result compared to established magnetic resonance spectroscopy (MRS)? 2. Is the myocardial fat-fraction obtained by VARPRO reproducible? The results suggest the two methods are strongly associated but agreed poorly. The VARPRO estimation is proportionally biased, however, the bias can be easily accounted for using a linear equation. The temporal reproducibility of VARPRO is satisfactory. VARPRO could be a potential, alternative modality to MRS in myocardial lipid quantification.

Richard Joshua Murdock^1,2 and Michael Cima^2,3
1Health Sciences & Technology, Massachusetts Institute of Technology, Cambridge, MA, United States, ²Koch Institute For Integrative Cancer Research, Cambridge, MA, United States, ³Materials Science & Engineering, Massachusetts Institute of Technology, Cambridge, MA, United States

Continuous biochemical monitoring remains a critical challenge in the biosensing community as longevity, biocompatibility, and signal transduction limit multi-month implantation. We propose a novel design and evaluation of the sensing capability of a magnetic nanoparticle dosimeter assay for in vivo implementation, readable by magnetic relaxation. This local diagnostic device offers continuous sentinel evaluation and longitudinal tracking of critical biomarkers found to correlate with cardiovascular disease states. The current design exhibits an order of magnitude improvement in performance compared to prior studies, translating into a predicted 29 week lifetime.

Dongyue Si¹, Yanfang Wu², Jie Yin², Rui Guo³, Jingjing Xiao⁴, Bowei Liu¹, Xue Lin², Peng Gao², Deyan Yang², Quan Fang², Jianwen Luo¹, Daniel A. Herzka^5,6, and Haiyan Ding^1
1Center for Biomedical Imaging Research (CBIR), Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, China, ²Department of Cardiology, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, China, ³Department of Medicine, Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA, United States, ⁴Department of Medical Engineering, Xinqiao Hosptial, Army Medical University, Chongqing, China, ⁵Department of Biomedical Engineering, Johns Hopkins School of Medicine, Baltimore, MD, United States, ⁶Cardiovascular Interventional Program, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, MD, United States

Radiofrequency ablation (RFA) is a first-line treatment for controlling atrial fibrillation (AF). Visualizing fibrotic tissues as well as the surrounding anatomical structure could improve the delivery of therapy. In this work, an independent navigator-gated three-dimensional dark-blood phase-sensitive inversion recovery sequence was developed for atrial visualization. Dark-blood was achieved by optimizing inversion recovery and T₂-preparation timing. Preliminary results from in vivo imaging in patients with AF post RFA demonstrate that the proposed technique can clearly visualize the atrial wall and scar formation. Both dark blood contrast and high resolution (1.25×1.25×3 mm³) were achieved. 

Robert J Holtackers^1,2,3, Suzanne Gommers^1,2, Caroline M van de Heyning^3,4, Jouke Smink⁵, Amedeo Chiribiri³, Joachim E Wildberger^1,2, and Rachel ter Bekke^2,6
1Department of Radiology & Nuclear Medicine, Maastricht University Medical Centre, Maastricht, Netherlands, ²CARIM School of Cardiovascular Diseases, Maastricht University, Maastricht, Netherlands, ³School of Biomedical Engineering and Imaging Sciences, King’s College London, London, United Kingdom, ⁴Department of Cardiology, Antwerp University Hospital, Antwerp, Belgium, ⁵Philips Healthcare, Best, Netherlands, ⁶Department of Cardiology, Maastricht University Medical Centre, Maastricht, Netherlands

Conventional 2D LGE often suffers from poor scar-to-blood contrast and limited spatial resolution. While 3D methods are known for their superior spatial resolution, our earlier proposed dark-blood LGE without additional magnetization preparation enables improved scar-to-blood contrast. In the present study we sought to assess the effect of a slowly increasing, dynamic inversion time on left-ventricular blood-pool suppression in a high-resolution isotropic 3D dark-blood LGE acquisition, and compare it against a conventional, fixed inversion time. Our findings show that such a dynamic inversion time significantly improves blood-pool suppression and thereby maximizes the dark-blood contrast mechanism for improved detection of myocardial scarring.

Gador Canton¹, Hiroko Watase², Josh Liu³, Li Chen⁴, Yin Guo⁴, Yongjun Wang⁵, and Chun Yuan^1
1Radiology, University of Washington, Seattle, WA, United States, ²Surgery, University of Washington, Seattle, WA, United States, ³University of Washington, Seattle, WA, United States, ⁴Electrical and Computer Engineering, University of Washington, Seattle, WA, United States, ⁵Beijing Tiantan Hospital, Capital Medical University, Beijing, China

This study aims to characterize the luminal morphology of middle cerebral arteries affected by intracranial atherosclerotic disease in a cohort of ischemic stroke patients. We used a visualization method (MOCHA) based on multiplanar reconstruction of the intracranial vasculature. We identified three distinct morphological patterns based on the effect of the atherosclerotic plaque presence on the downstream geometry. Together with the location of the plaque with respect to the middle cerebral artery branching point from the internal carotid artery, these distinct morphological patterns might be the differential factor in recurrent stroke risk stratification due to the resulting local hemodynamics. 

Liam Timms¹, Tianyi Zhou¹, Ju Qiao², Vishala Mishra³, Gayatri Veeramani^4,5, Andrew Allegretti⁴, Thomas Benkert⁶, John Kirsch², Ravi Seethamraju⁷, Mukesh Harisinghani³, and Srinivas Sridhar^1,8
1Physics, Northeastern University, Boston, MA, United States, ²Massachusetts General Hospital, Boston, MA, United States, ³Radiology, Massachusetts General Hospital, Boston, MA, United States, ⁴Nephrology, Massachusetts General Hospital, Boston, MA, United States, ⁵Nephrology, Brigham and Women’s Hospital, Boston, MA, United States, ⁶Siemens Healthcare, Erlangen, Germany, ⁷Siemens Healthcare, Boston, MA, United States, ⁸Theranano LLC, Newton, MA, United States

A contrast-enhanced MRA technique using Ultra-short Time-to-Echo (UTE) pulse sequence with super-paramagnetic iron oxide nanoparticle (SPION) contrast agent is demonstrated for abdominal vascular imaging in human patients. It successfully depicts the kidney anatomy, as well as kidney cysts and kidney vascular structure in excellent clarity and detail. With manual segmentation, total kidney volume, cyst volume and renal artery diameter were estimated. This initial study in humans demonstrated the basic feasibility of abdominal QUTE-CE MRA as an alternative to Gd based CEMRA.

Javier Montalt-Tordera¹, Grzegorz Kowalik¹, Alexander Gotschy^2,3, Jennifer Steeden¹, and Vivek Muthurangu^1
1Institute of Cardiovascular Science, UCL, London, United Kingdom, ²Great Ormond Street Hospital, London, United Kingdom, ³Institute for Biomedical Engineering, ETH Zurich, Zurich, Switzerland

A free-breathing three-dimensional cine imaging sequence was developed using a stack of spirals trajectory with golden angle rotations for efficient k-space sampling. Parallel imaging and compressed sensing were used in reconstruction for data acceleration. Respiratory motion artefact was avoided by binning k-space data into multiple respiratory phases. The proposed imaging protocol was tested on 10 patients and compared with a standard 2D breath-held short-axis stack. The proposed method was faster to acquire and generally able to provide correct quantitative measurements, but image quality needs to be improved and reconstruction time is too long for clinical practice.

W. Patricia Bandettini¹, Sujata M. Shanbhag¹, Christine M. Mancini¹, Delaney R. McGuirt¹, Jennifer L. Henry¹, Margaret M. Lowery¹, Marcus Y. Chen¹, Hui Xue¹, Peter M. Kellman¹, and Adrienne E. Campbell-Washburn^1
1NATIONAL INSTITUTES OF HEALTH/NHLBI, BETHESDA, MD, United States

We demonstrate the diagnostic capabilities of a high-performance, low-field (0.55 Tesla) scanner in the acquisition and interpretation of late gadolinium enhancement (LGE) in patients referred for assessment of myocardial infarction (MI).  Patients underwent paired comparison exams with breath-held gradient echo LGE imaging at 1.5T and breath-held bSSFP LGE imaging at 0.55T. The number of enhancing segments identified between each field strength was similar (59 segments at 0.55T vs 63 segments at 1.5T), and assessment of epicardial coronary artery distribution matched exactly between the two field strengths; included were two multi-vessel disease cases.

Fadil Ali¹, Mark Bydder¹, Da Wang², Vahid K Ghodrati¹, Chang Gao¹, Ashley Propser¹, Kim-Lien Nguyen¹, and Peng Hu^1
1Radiology, University of California, Los Angeles, Los Angeles, CA, United States, ²Radiology, The Unversity of Chicago, Chicago, IL, United States

Balanced steady-state free precession (bSSFP) is susceptible to outflowing signal because of it features no spoiling. This results in misregistrated artifacts that hamper the cardiac evaluations.  We propose a method that tackles this issue in the acquisition process by spatially encoding for this outflowing signal to prevent it from projecting onto the imaged slice.  

Anna Padée¹, Lorenzo Di Sopra², John Heerfordt^2,3, Jérôme Yerly^2,4, Marco Merlo¹, Tobias Kober^2,3,5, Davide Piccini^2,3,5, Matthias Stuber^2,4, Christopher Roy², and Jonas Richiardi^2,3
1Laboratory for Psychiatric Neuroscience and Psychotherapy, University of Fribourg, Fribourg, Switzerland, ²Department of Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, ³Advanced Clinical Imaging Technology, Siemens Healthcare, Lausanne, Switzerland, ⁴Center for Biomedical Imaging, Lausanne, Switzerland, ⁵LTS5, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland

ECG recording during whole heart MR imaging requires extra setup time and is not always reliable. Recently, self-gated (SG) acquisition techniques have been developed, which extract the cardiac signal from k-space center amplitude modulations. Here, we investigated whether different single- (Empirical Mode Decomposition (EMD), Ensemble EMD (EEMD)) and multi-coil (PCA, ICA with novel stabilisation) decomposition methods yielded SG triggers with minimal variability to ECG R-wave location. EEMD yielded median variability similar to PCA. Compared to PCA, our proposed stabilised ICA approach yielded SG triggers with 66% lower variability (median over 5 subjects), although not for all subjects.

Delphine Perie¹, Mehdi Ghafari¹, and Daniel Curnier^2
1Mechanical Engineering, Polytechnique Montreal, Montreal, QC, Canada, ²Kinesiology, University of Montreal, Montreal, QC, Canada

Due to the complexity of the heart’s geometry, the characterization of the material properties of the myocardium remains challenging. Using kinematically admissible hyperboloid virtual fields and a finite element mesh of the left ventricle, we successfully solved the virtual field method equations to determine the stiffness of the myocardium from a reverse identification approach. The distribution of the elastic stiffness presented same patterns between the 3 volunteers. However, the standard risk volunteer presented higher stiffness within the mobile wall than the high risk volunteers. This approach would be an efficient tool to characterize early cardiac dysfunction.

Bonnie Lam¹, Barry Fung¹, and Moriel Vandsburger^1
1University of California, Berkeley, Berkeley, CA, United States

While accelerated imaging via compressed sensing reconstruction has been explored in several dynamic contrast settings, incorporation of compressed sensing for accelerated image acquisition has not been fully explored in the setting of cardiac CEST. Here we test the impact of various compressed sensing methods and acceleration factors upon the accuracy of endogenous cardiac CEST in mice, specifically measures of specificity, accuracy, and sensitivity.

Samuel Edward Bibelhauser¹, Sean P Callahan¹, Narayana Singam², Marcus Stoddard^2,3, and Amir Amini^1,3
1Electrical and Computer Engineering, University of Louisville, Louisville, KY, United States, ²Division of Cardiovascular Medicine, University of Louisville, Louisville, KY, United States, ³Radiology, VA Medical Center, Louisville, KY, United States

Effective Orifice Area (EOA) is a valuable diagnostic parameter for the determination of the severity of Aortic Stenosis(AS). A 4D Flow based measurement of this parameter was developed, and tested.  The method was validated on a rigid phantom, applied to 4 healthy volunteers and 17 patient subjects with moderate to severe AS.  The resultant EOA, for patients, was compared to their transthoracic echocardiography (TTE) with Doppler result.  The comparison used Bland Altman analysis and linear regression which revealed a bias of 0.04 cm² and correlation coefficient of 0.92. Showing a high level of concordance between the two modalities.

Sujata M. Shanbhag¹, Rajiv M. Ramasawmy¹, W. Patricia Bandettini¹, Christine M. Mancini¹, Delaney R. McGuirt¹, Jennifer L. Henry¹, Margaret M. Lowery¹, Marcus Y. Chen¹, and Adrienne Campbell-Washburn^1
1NATIONAL INSTITUTES OF HEALTH/NHLBI, BETHESDA, MD, United States, Bethesda, MD, United States

Low-field MRI equipped with high-performance technology and modern imaging methods offers the potential to make technically demanding cardiac imaging more accessible. This study demonstrates that a low-field (0.55T) system can provide accurate phase-contrast flow measurements in both healthy volunteers and patients who were clinically referred for assessment of intracardiac shunt and valvular disease. Comparative studies between contemporary high-performance 0.55T and 1.5T systems generated well-correlated quantitative flow measurements (r>0.87). 

Simon Schmidt¹, Kristine John², Martin Bruschewski², Sebastian Flassbeck¹, Mark E. Ladd¹, Sven Grundmann², and Sebastian Schmitter^1,3
1Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany, ²Institute of Fluid Mechanics, University of Rostock, Rostock, Germany, ³Physikalisch-Technische Bundesanstalt (PTB), Braunschweig and Berlin, Germany

In this work, we successfully quantified the Reynolds stress tensor (RST) in a well-known fluid-dynamic test case: the flow over periodic hills. This test case was chosen because the turbulence in this kind of flow is strongly inhomogeneous and anisotropic, representing a challenging measurement task. The results indicate, in analogy to intravoxel velocity standard deviation (IVSD) mapping, that RST quantification is highly susceptible to the applied $$$m_1^{enc}~$$$value. Furthermore, RST mapping inherently requires a higher dynamic range compared to IVSD mapping, since the shear stresses are typically much lower than the normal stresses.

Jost Michael Kollmeier¹, Dirk Voit¹, and Jens Frahm^1
1Biomedizinische NMR, Max-Planck-Institut für biophysikalische Chemie, Göttingen, Germany

Real-time phase-contrast flow MRI based on highly undersampled radial FLASH employs fast gradients that make concomitant field contributions relevant even at 3 T.  In order to correct for phase errors a spoke-wise Maxwell correction has been developed and successfully integrated into a model-based reconstruction. For radial trajectories corrections for individual k-space lines are required as concomitant fields may change from spoke to spoke. The proposed method significantly improves phase errors in velocity maps as shown for three-directional real-time phase-contrast flow MRI of the human aortic arch.

Dan Zhu^1,2 and Michael Schär^2
1Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, United States, ²Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, United States

Velocity encoded MRI has clinical significance in the diagnosis of coronary artery disease. High spatial and temporal resolution increase the accuracy and reproducibility and preserve reliable flow pattern. By using rotated golden angle with k-t space reconstruction and retrospective cardiac gating, high quality flow CINE could be achieved with high spatial and temporal resolution during a single breath-hold. However, binning to cardiac frames may lead to large gaps in k-space. In this work, we propose a new scheme of rotated golden angle and optimized retrospective gating by shifting the binning window. Improved image quality is demonstrated with the proposed methods.

Hannes Dillinger¹, Jonas Walheim¹, Robbert J.H. van Gorkum¹, and Sebastian Kozerke^1
1University and ETH Zurich, Zurich, Switzerland

An evaluation of velocity field reconstructions from echo-planar trajectories in PC-MRI relative to ground truth computational fluid dynamics (CFD) data in stenotic high flow regimes is presented. The reconstructed velocity images and variations in spatial resolution resulting from modulation of the Point Spread Function (PSF) are investigated. Significant artifacts and resolution loss of up to factor 8 for EPI in high-flow regimes of 220cm/s are found. In-vitro experiments confirm that depending on the orientation of the main flow direction the reconstructed velocity images differ significantly in regions of high shear or acceleration with implications for in-vivo applications.

Taehoon Shin^1,2 and Wanyong Shin^3
1Ewha Womans University, Seoul, Korea, Republic of, ²Case Western Reserve University, Cleaveland, OH, United States, ³Cleveland Clinic, Cleveland, OH, United States

Based on the magnitude similarity between bipolar-encoded k-space data for PC flow imaging, magnitude-difference regularization was incorporated into the conventional compressed sensing (CS) reconstruction. The gradient of the magnitude regularization was derived so the reconstruction problem can be solved iteratively. Retrospecitve in-vivo studies show that the addition of magnitude-difference regularization into conventional CS reconstruction improves the accuracy of image reconstruction using highly undersampled phase-contrast flow MR data. 

Grant S Roberts¹, Kevin M Johnson^1,2, Steven R Kecskemeti², Ozioma Okonkwo³, Sarah Lose³, Laura Eisenmenger², and Oliver Wieben^1,2
1Medical Physics, University of Wisconsin - Madison, Madison, WI, United States, ²Radiology, University of Wisconsin - Madison, Madison, WI, United States, ³Medicine, University of Wisconsin - Madison, Madison, WI, United States

Pulse wave velocity (PWV) is a biomarker that indirectly relates to arterial stiffness, an early indicator of cardiovascular disease. Breath-hold phase contrast (PC) MRI can be used to assess PWV in the aorta, however in certain populations, breath-holds may be difficult. We present a method to measure aortic PWV using a free-breathing radially-undersampled PC sequence. Initial results show that the free-breathing PC-derived PWV measures are comparable to the measures obtained from breath-hold Cartesian PC scans. Larger cohorts are warranted to verify these findings.  

Takashi Mizuno¹, Yasuo Takehara², Masataka Sugiyama², Ryota Horiguchi³, Shinji Naganawa³, Yasuo Sakurai¹, Yutaka Kato¹, Shinji Abe¹, Haruo Isoda⁴, Tomohiro Sato⁵, Tsuneo Ishiguchi⁵, Masanori Tadokoro⁵, and Atushi Nozaki^6
1Department of Radiological Technology, Nagoya University Hospital, Nagoya, Japan, ²Department of Fundamental Development of Low Invasive Diagnostic Imaging, Nagoya University Graduate School of Medicine, Nagoya, Japan, ³Department of Radiology, Nagoya University Graduate School of Medicine, Nagoya, Japan, ⁴Brain & Mind Research Center, Nagoya University, Nagoya, Japan, ⁵Trust Clinic, Nagoya, Japan, ⁶GE Healthcare Japan, Tokyo, Japan

We analyzed energy loss in normal subjects, patients in pre- and post- endovascular aortic repair. Patients had vortices in dilated flow path in pretreatment and tortuous high-velocity streamlines in posttreatment. Energy loss of posttreatment was higher than that of pretreatment. Energy loss would be affected most effectively by the change of the flow direction.

Zhensen Chen¹, Li Chen², Manabu Shirakawa¹, Wenjin Liu¹, Dakota Ortega¹, Jinmei Chen¹, Niranjan Balu¹, Theodore Trouard³, Thomas S Hatsukami⁴, Wei Zhou⁵, and Chun Yuan^1
1Radiology, University of Washington, Seattle, WA, United States, ²Electrical and computer engineering, University of Washington, Seattle, WA, United States, ³Biomedical Engineering, University of Arizona, Tuscon, AZ, United States, ⁴Surgery, University of Washington, Seattle, WA, United States, ⁵Surgery, University of Arizona, Tuscon, AZ, United States

In this study, we assessed the potential of using vasculature features extracted from TOF images as surrogate markers of intracranial blood flow in patients undergoing carotid revascularization surgery and MR imaging prior to, within 48 hours after and 6 months after surgery. The intracranial vasculature features, including total volume, total length and total number of branches etc., on TOF images were extracted using a novel tool named iCafe, and then compared with arterial spin labeling (ASL) cerebral blood flow (CBF) measurement. The results show that TOF-iCafe vasculature features have similar behaviors as ASL CBF over the time points.

Yuli HUANG¹, Li ZHU², Haiyang DONG², Yuan QU³, and Yan WANG^3
1MR, Philips Healthcare (Suzhou) Co., Ltd, SUZHOU, China, ²Radiology, Shanghai Chest Hospital, Shanghai JiaoTong University, Shanghai, China, ³Radiology, People’s Hospital of Xinjiang Uygur Autonomous Region, Urumqi, China

4D Flow MRI is a technique that can characterize vascular structure and hemodynamics. Compared with SENSE, Compressed SENSE can further reduce scanning time using the sparse characteristics of MR images. This study aims to investigate the feasibility of 4D Flow MRI with Compressed SENSE by comparing with 4D Flow MRI with SENSE. 4D Flow dynamic angiography of hepatic portal vein using SENSE and Compressed SENSE could be completed within 5 minutes with optimized parameters. The scanning time of Compressed SENSE 4D Flow was shorter than that of SENSE 4D Flow.

Jerome Lamy¹, Gigi Galiana², and Dana Peters^2
1Department of Radiology and Biomedical Imaging, Yale University, New Haven, CT, United States, ²Department of Radiology and Biomedical Imaging, Yale, New Haven, CT, United States

Phase-contrast (PC) MRI requires at least doubled scan time compared to anatomical imaging. This has required careful trade-offs regarding spatial and temporal resolution, and breath-holding times, and often the temporal or spatial resolution is insufficient.  Here we introduce a simple novel method which temporally under-samples acquisition of the phase-reference images, providing increased temporal resolution. This method is evaluated both in the aortic and pulmonary vein flow, which exhibits several rapid flow peaks requiring good spatial resolution. This method allowed to accurately identify velocity peaks for the aorta and pulmonary veins, providing higher temporal resolution as compared to the standard PC-MRI.

Min Liu¹, Xincao Tao², Jing An³, Ning Jin⁴, and Fan Lin^5
1Radiology, China-Japan Friendship Hospital, Beijing, China, ²Department of Pulmonary and Critical Care Medicine, China-Japan Friendship Hospital, Beijing, China, ³Siemens Shenzhen Magnetic Resonance Ltd, Siemens Shenzhen Magnetic Resonance Ltd, Shen Zhen, China, ⁴Siemens Medical Solution, Siemens Medical Solution, Chicago, IL, United States, ⁵Department of Cardiovascular surgery, China-Japan Friendship Hospital, Beijing, China

Using 4D FLOW CMR, flow components including percent of direct flow (PDF), retained inflow (PRI), delayed ejection flow (PDE) and residual volume (PRV) of bi-ventricles in patients with PH were quantified. Compared with non-PH patients, flow components had significantly alterations. RV PDF and PRV significantly correlated with RVEDV, RSV, RVEF and RV mass as well as NT-proBNP. Moreover, RV PDF and PVR correlated with pulmonary artery pressure and PVR. We think that 4D flow CMR may become a powerful surrogate for the future PH studies

Dominik Daniel Gabbert¹, Arash Kheradvar², Michael Jerosch-Herold³, Thekla Oechtering⁴, Felix Wadle¹, Anselm Sebastian Uebing¹, Hans-Heiner Kramer¹, Inga Voges¹, and Carsten Rickers^1
1Department of Congenital Heart Disease and Pediatric Cardiology, University Hospital Schleswig-Holstein, Kiel, Germany, ²The Edwards Lifesciences Center for Advanced Cardiovascular Technology, University of California, Irvine, CA, United States, ³Harvard Medical School, Brigham and Women’s Hospital, Boston, MA, United States, ⁴Department of Radiology and Nuclear Medicine, University Hospital Schleswig-Holstein, Lübeck, Germany

Many secondary flow patterns in the large vessels have been shown to represent significant pathophysiological phenomena. However, the computational cost to analyze a large number of quantities usually discourage the use of these parameters for clinical decision making given the limited computing resources in clinical settings. We describe a novel postprocessing method for comprehensive analysis of vascular anatomy and fluid dynamics based on 4D Flow MRI. The method defines a multi-dimensional feature space built from a few complementary fluid dynamics building blocks to efficiently determine a large number of anatomic and fluid dynamics parameters in the course of a vessel.

Jeesoo Lee^1,2, Nadia El Hangouche², Alex J Barker³, James D Thomas², and Michael Markl^1,4
1Radiology, Northwestern University, Chicago, IL, United States, ²Cardiology, Northwestern Memorial Hospital, Chicago, IL, United States, ³Radiology, University of Colorado Denver, Denver, CO, United States, ⁴Biomedical Engineering, Northwestern University, Chicago, IL, United States

We developed a framework to integrate 4D flow MRI with color Doppler echocardiography (CDE) for direct localized inter-modal velocity comparison. The framework combines CDE and 4D flow MRI data by using cine-MRI and a two-step image registration; 1) registration of CDE to cine-MRI and 2) registration of cine-MRI to 4D flow MRI. As a result, a voxel-wise inter-modal velocity difference map overlaid to a tissue image was constructed. Six healthy subjects with a same-day CDE and 4D flow MRI were utilized and a strong correlation with moderate agreement was found in the LVOT.

Mariana B. L. Falcão¹, Lorenzo Di Sopra¹, Liliana Ma^2,3, Mario Bacher^1,4, Davide Piccini^1,5, Jérôme Yerly^1,6, Peter Speier⁴, Tobias Rutz⁷, Milan Prša⁷, Michael Markl^2,3, Matthias Stuber^1,6, and Christopher Roy^1
1Department of Diagnostic and Interventional Radiology, Lausanne University Hospital (CHUV) and University of Lausanne (UNIL), Lausanne, Switzerland, ²Department of Radiology, Feinberg School of Medicine, Northwestern University, Chicago, IL, United States, ³Department of Biomedical Engineering, Northwestern University, Chicago, IL, United States, ⁴Siemens Healthcare GmbH, Erlangen, Germany, ⁵Advanced Clinical Imaging Technology, Siemens Healthcare AG, Lausanne, Switzerland, ⁶Center for Biomedical Imaging, Lausanne, Switzerland, ⁷Department of Cardiology, Lausanne University Hospital (CHUV), Lausanne, Switzerland

Conventional 4D flow MRI techniques often have prolonged and unpredictable scan times, due to the use of respiratory navigation. To address this, a fully self-gated cardiac and respiratory motion-resolved whole-heart 5D flow protocol with a fixed scan time was recently developed using a free-running framework. This protocol extracts cardiac and respiratory signals from periodic readouts (self-gating). In this study, we explore the use of Pilot Tone signals as an alternative method for cardiac and respiratory signal extraction to reconstruct 5D flow data and compare reconstructions to those using the previously established self-gating method and the conventional 4D flow sequence.

Philip A Corrado¹, Rafael Medero², Kevin M. Johnson^1,3, Christopher J François ³, Alejandro Roldán-Alzate^2,4, and Oliver Wieben^1,3
1Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, ²Mechanical Engineering, University of Wisconsin-Madison, Madison, WI, United States, ³Radiology, University of Wisconsin-Madison, Madison, WI, United States, ⁴Biomedical Engineering, University of Wisconsin-Madison, Madison, WI, United States

Dual-Venc 4D flow MRI offers higher velocity sensitivity than single-Venc imaging but requires increased scan time. Radial sampling, either with a 3D-radial or hybrid stack-of-stars (SOS) trajectory, allows for acceleration that can offset the increase, but the optimal technique depends on anatomical region and is not apparent for imaging the left ventricle (LV). We compared 3D-radial and SOS velocity images in an in-vitro LV model versus a reference dataset acquired with particle imaging velocimetry (PIV), an experimental optical imaging technique. 3D-radial matched PIV velocities better than SOS, suggesting it is better for accelerated, dual-Venc imaging of the LV.

Takashi Fujiwara¹, Lorna Browne¹, Ladonna Malone¹, Quin Lu², Brian Fonseca³, Michael DiMaria³, and Alex J Barker^1,4
1Department of Radiology, Section of Pediatric Radiology, Children's Hospital Colorado, University of Colorado Anschutz Medical Campus, Aurora, CO, United States, ²Philips Healthcare NA, San Francisco, CA, United States, ³Department of Pediatrics, Section of Pediatric Cardiology, Children's Hospital Colorado, University of Colorado Anschutz Medical Campus, Aurora, CO, United States, ⁴Department of Bioengineering, University of Colorado Anschutz Medical Campus, Aurora, CO, United States

We investigate the feasibility of the accelerating 4D flow with compressed sensing and online reconstruction in pediatric patients. In-vitro models and imaging in pediatric patients was performed and hemodynamic quantification was compared to conventional approaches. The results demonstrated flow metrics and clinically relevant indices measured by 6-fold accelerated CS were in good agreement with those measured by conventional acceleration techniques, with scan time savings of 30.2%. This suggests the feasibility of using CS 4D flow for pediatric patients in the clinical setting.

Qingdi Wang¹, Xiaojing Guo¹, Emma Hornsey², Lucy McKenna², Leonid Churilov^1,3, Mark Brooks^1,2, George Matalanis^1,2, Jason Chuen^1,2, Eric Poon¹, Daniel Staeb⁴, Ning Jin⁵, Andrew Ooi¹, and Ruth P Lim^1,2
1The University of Melbourne, Melbourne, Australia, ²Austin Health, Heidelberg, Australia, ³The Florey Institute of Neuroscience and Mental Health, Melbourne, Australia, ⁴Siemens Healthcare Pty Ltd, Melbourne, Australia, ⁵Siemens Medical Solutions, Chicago, IL, United States

A four-dimensional phase-contrast magnetic resonance imaging sequence with respiratory-controlled adaptive k-space reordering (ReCAR-4DPC) offers potential benefits of improved scan efficiency and motion robustness.  Imaging was performed on 15 volunteers and 2 patients with aortic dissection. Inter-scan repeatability and inter-reader agreement of flow metrics derived from ReCAR-4DPC was assessed and compared to metrics derived from 2-dimensional phase contrast imaging (2DPC). There was almost perfect inter-scan and inter-reader concordance (Lin's concordance correlation coefficient for all metrics > 0.91 and >0.98 respectively). Concordance with 2DPC was also high (LCCC all > 0.873). RC-4DFlow is reproducible and repeatable, with high concordance with 2DPC metrics.

Stefano Buoso¹, Hannes Dillinger¹, and Sebastian Kozerke^1
1Institute for Biomedical Engineering, ETH Zurich, Zurich, Switzerland

We propose a variational method for reconstructing divergence-free velocity fields from 4D Flow MRI data with compensation of spatial displacement artifacts. A standard  4D Flow MRI sequence to measure flow velocities in an expanding flow phantom with Reynolds number of 4000 was used. A variational formulation allows to reconstruct a divergence-free velocity field used to estimate the displacement artefact associated with spin motion between the time point of motion encoding and echo time. The field was used to apply a spatial shift of the original MRI data and compute new divergence-free flow velocities corrected for spatial motion artifacts.

Corina Kräuter^1,2, Ursula Reiter¹, Clemens Reiter¹, Volha Nizhnikava¹, Marc Masana³, Albrecht Schmidt⁴, Michael Fuchsjäger¹, Rudolf Stollberger², and Gert Reiter^5
1Department of Radiology, Medical University of Graz, Graz, Austria, ²Institute of Medical Engineering, Graz University of Technology, Graz, Austria, ³Computer Vision Center, Universitat Autònoma de Barcelona, Barcelona, Spain, ⁴Department of Internal Medicine, Medical University of Graz, Graz, Austria, ⁵Research and Development, Siemens Healthcare Diagnostics GmbH, Graz, Austria

Automated mitral valve vortex ring evolution analysis from magnetic resonance 4D flow data is feasible. However, time-consuming manual segmentation of the left ventricular blood pool represents a bottleneck. We simulated speed-up and variability of manual segmentation and analyzed the impact on vortex ring parameters. Automated mitral valve vortex ring extraction and analysis yielded robust results, even when applying the end-diastolic segmentation mask to all cardiac phases. Error analysis of vortex ring parameters showed that under-segmentation of the ventricular blood pool should be avoided. Speeding up segmentation by using only the end-diastolic mask enables clinical mitral valve vortex ring analysis studies.

Wen Chen¹, Song Yang¹, Qian Lu¹, Lin Xu¹, Yang Fan², Yong Zhang³, and Changjie Pan^3
1Radiology department, Hubei Shiyan Taihe Hospital, Shiyan, China, ²GE healthcare China, Beijing, China, ³Radiology department, Changzhou No2 people hospital, Changzhou, China

Respiratory gating or navigator technique is used in 4D flow imaging to reduce the effect of respiratory motion. Non-respiratory gated 4D flow MRI reduces the overall scan time but may face uncertainty in reproducibility and reliability. In this work, the reliability and reproducibility of non-respiratory gated 4D flow was investigated in 10 patients and 3 different sites in two separate scans. Better measurement consistency was observed intra-site than inter-site. The intra-correlation coefficient results showed excellent to good reproducibility between sites were observed ( ICC > 0.95 for blood flow, ICC > 0.85 for peak).  Bland–Altman analysis also showed great test-retest reliability. 

Henrik Haraldsson¹, Keerthi Valluru¹, Megan Ballweber¹, Ning Jin², Sinyeob Ahn³, Evan Kao¹, David Saloner^1,4, and Matthew Amans^1
1University of California, San Francisco, San Francisco, CA, United States, ²Siemens Healthcare, Chicago, IL, United States, ³Siemens Healthcare, San Francisco, CA, United States, ⁴Veterans Affairs Medical Center, San Francisco, CA, United States

Idiopathic intracranial hypertension is a condition of elevated intracranial cerebrospinal fluid pressure. Transverse sinus stenoses are clearly related to the pathophysiology, and catheter-based manometry is the “gold standard” to assess pressure gradients across these stenoses. The aim of this work is to evaluate the feasibility of assessing these pressure gradients in the intracranial veins using 4DFlow with ICOSA6 motion encoding – which also account for turbulence - and to compare these pressure gradients with catheter-based manometry.  Pressure gradient were measured in patient specific phantoms using both 4DFlow and catheter-based manometry.  The pressure gradient assessed showed good agreement when accounting for turbulence. 

Julio Sotelo^1,2,3, Malenka M Bissell⁴, Yaxin Jiang⁴, Hernan Mella^1,2,3, Joaquin Mura^3,5, and Sergio Uribe^1,3,6,7
1Biomedical Imaging Center, Pontificia Universidad Catolica de Chile, Santiago, Chile, ²Department of Electrical Engineering, Pontificia Universidad Catolica de Chile, Santiago, Chile, ³Millennium Nucleus for Cardiovascular Magnetic Resonance, Santiago, Chile, ⁴Department of Biomedical Imaging Science, Leeds Institute to Cardiovascular and Metabolic Medicine, University of Leeds, Leeds, United Kingdom, ⁵Department of Mechanical Engineering, Universidad Técnica Federico Santa María, Santiago, Chile, ⁶Department of Radiology, Pontificia Universidad Catolica de Chile, Santiago, Chile, ⁷Institute for Biological and Medical Engineering, Schools of Engineering, Medicine and Biological Sciences, Pontificia Universidad Catolica de Chile, Santiago, Chile

The circulation is normally analyzed in a 2D cross-section of the aorta, manually placed. To avoid this problem, evaluate and validate a new methodology based on Finite Elements (FE) to calculate the circulation in three-dimensions, in in-silico models, and in the ascending aorta of a group of volunteers and patients. In in-silico experiment we obtain an RMSE of the circulation values less than 1.6e^-6. We also found that significant differences exist between volunteers and patients with a p-value of 0.0283.  Our method is straight forward to calculate, accurate and robust throughout different resolutions and noise levels.

Lixing Ren^1,2, Xiaowei He², Dandan Zheng³, and Enhua Wu^2,4
1University of Chinese Academy of Sciences, Beijing, China, ²The State Key Lab. of CS, Institute of Software, Chinese Academy of Sciences, Beijing, China, ³Clinical Science, Philips Healthcare, Beijing, China, ⁴University of Macau, Macau, China

Understanding the nature of the pressure changes is crucial for diagnosis and therapy strategy optimization for aortic aneurysm. 4D flow MRI had offered the opportunity to assess 3D blood flow characteristics. However, the pressure quantification is still a challenge due to the low signal-to-noise ratio and the partial volume effects near the wall. The focus of this work is to address this deficiency by proposing a novel workflow of direct pressure estimation based on  simulated noisy 4D flow MRI data.

Judith Zimmermann^1,2,3, Kathrin Bäumler¹, Michael Loecher^1,3, Alison Marsden^4,5, and Daniel Ennis^1,3,5
1Radiology, Stanford University, Stanford, CA, United States, ²Computer Science, Technical University of Munich, Munich, Germany, ³Radiology, Veterans Administration Health Care System, Palo Alto, CA, United States, ⁴Pediatrics, Stanford University, Stanford, CA, United States, ⁵Cardiovascular Institute, Stanford, CA, United States

A detailed understanding of flow dynamics in the aorta is of clinical interest and enabled by 4D-flow MRI. This work presents an in vitro flow circuit that embeds a subject-specific 3D-printed compliant aorta model to demonstrate feasibility while also exploring the impact of vessel wall characteristics (thickness, stiffness) and heart rate on quantitative flow dynamics. The experimental setup will be a valuable tool for assessing 4D flow MRI sampling requirements, generating high-quality ground truth data for CFD and FSI validation, as well as studying flow dynamics in different vascular pathologies.