View Presentation Video

Pietro Dirix¹, Stefano Buoso¹, and Sebastian Kozerke^1
1University and ETH Zurich, Zürich, Switzerland

Keywords: Flow, Image Reconstruction

We expand a previously developed pipeline for generation of synthetic 4D flow MRI of turbulent flows to account for realistic k-space filling and add an MR temporal averaging component to the simulations. We show that temporal averaging in typical undersampled 4D flow MRI is imperfect, leading to inherent scan-to-scan variations of measured turbulence fields. We demonstrate how encoding schemes, when designed appropriately, allow to significantly reduce scan-to-scan variability, increasing the confidence on MRI measurements. 

View Presentation Video

Amirkhosro Kazemi¹, Marcus Stoddard¹, and Amir A. Amini^1
1Electrical and Computer Engineering, University of Louisville, Louisville, KY, United States

Keywords: Phantoms, Heart, Hemodynamics, Flow, Neural network, Deep learning

Variations in velocity derivative fluctuations have been correlated with changes in pressure gradient. Image denoising and super-resolution techniques are required to accurately quantify velocity fluctuations. We propose a novel network, which we call TKE-Net to estimate Turbulent Kinetic Energy (TKE) which utilizes a ResNet convolutional neural network backbone. The network is trained and tested with low-resolution simulated CFD data, as could be derived from low resolution 4D flow in a phantom model of arterial stenosis. The results indicate good accuracy in estimating TKE. The method was also applied to in-vitro 4D flow MRI data in identical geometry. 

View Presentation Video

Charles McGrath¹, Pietro Dirix¹, Jonathan Weine¹, and Sebastian Kozerke^1
1University and ETH Zurich, Zurich, Switzerland

Keywords: Flow, Velocity & Flow, Turbulence

The accuracy of turbulent flow quantification using phase-contrast MRI depends on both the turbulence energy spectrum to be imaged and the sampling spectrum of the velocity encoding gradients. In the present work their interplay is studied in detail, revealing erroneous quantification of intravoxel standard deviation and turbulent kinetic energy in particular in cases exhibiting low turbulence frequencies. It is concluded that experimental design for turbulent flow quantification needs to consider the joint effect of velocity encoding and imaging gradients.

Keywords: Turbulence, TKE, Phase Constrast, Simulation, Flow, Cardiovascular

View Presentation Video

Jie Xiang¹, Jerome Lamy², Maolin Qiu², and Dana C. Peters^1,2
1Department of Biomedical Engineering, Yale University, New Haven, CT, United States, ²Department of Radiology and Biomedical Imaging, Yale University, New Haven, CT, United States

Keywords: Flow, Quantitative Imaging

Diastolic function evaluation requires estimates of early and late diastolic mitral valve flow velocities (E and A), and mitral annulus tissue velocity (e’). Our goal was to develop a bSSFP phase-contrast (PC) sequence (PC-SSFP) for in-plane flow-encoding, for simultaneous recording of E and A, and estimation of e’ based on tracking mitral valve motion on cine magnitude images, in a single breath-hold. Phantom and in vivo experiments showed agreement of PC-SSFP with velocities on GRE-based PC providing similar velocity curves, while achieving the high SNR and contrast of bSSFP cine images.

View Presentation Video

Tarun Naren¹, Oliver Wieben^1,2, Thekla H Oechtering^2,3, Scott B Reeder^1,2, and Kevin M Johnson^1,2
1Department of Medical Physics, University of Wisconsin - Madison, Madison, WI, United States, ²Department of Radiology, University of Wisconsin - Madison, Madison, WI, United States, ³Department of Radiology, Universität zu Lübeck, Lübeck, Germany

Keywords: Flow, Liver

Radially undersampled 4D flow MRI is a promising method for non-invasive mapping of blood flow in the portal venous system. However, collecting sufficient projections to produce clinically viable images can lead to long scan times (10+ minutes) as fewer projections cause undersampling artifacts that appear as structured noise. In this study, we propose a data-driven, deep learning method to denoise vastly undersampled (<10% of full Nyquist sampling) radial 4D flow MRI data in the portal vein. We train a network on a heterogeneous, time-averaged dataset with two levels of undersampling and perform a quantitative hemodynamic analysis to compare results.

View Presentation Video

Julio A. Oscanoa^1,2, Matthew J. Middione², Michael Loecher², Ana Beatriz Solana³, Shreyas S. Vasanawala², and Daniel B. Ennis^2
1Department of Bioengineering, Stanford University, Stanford, CA, United States, ²Department of Radiology, Stanford University, Stanford, CA, United States, ³GE Healthcare, Munich, Germany

Keywords: Flow, System Imperfections: Measurement & Correction, Phase-Contrast, Eddy currents

We propose a physics-driven approach for prediction of eddy current-induced background phase offsets in 2D PC-MRI. The method leverages the applied gradient waveforms and a circuit model to estimate the system-dependent parameters that govern the background phase. The aim is to use these parameters and the model to predict the background phase of clinical acquisitions from pulse sequence information only. Our method achieved an RMSE of 1.05 cm/s and adequate visual quality.

View Presentation Video

Mahmoud Ebrahimkhani¹, Ethan Johnson¹, Aparna Sodhi², Joshua Robinson^2,3, Cynthia Rigsby², Bradly Allen³, and Michael Markl^3
1Radiology, Northwestern University, Chicago, IL, United States, ²Ann & Robert H. Lurie Children’s Hospital, Chicago, IL, United States, ³Northwestern University, Chicago, IL, United States

Keywords: Flow, Heart

We pursued a deep learning approach to investigate the utilization of a wearable seismocardiography (SCG) device to predict measures of flow similar to those obtained using 4D flow MRI. SCG can measure the chest vibrations caused by cardiac mechanical activities such as valve closures and changes of pulsatile flow. We hypothesized that deep learning can be used to infer the pathological changes in blood flow, such as a higher systolic peak velocity (Vmax) in patients with aortic valve diseases, from the SCG signals.

View Presentation Video

Matthew J. Middione¹, Julio A. Oscanoa^1,2, Xianglun Mao³, Christina R. Ruiz⁴, Michael Salerno^1,4,5, Shreyas S. Vasanawala¹, and Daniel B. Ennis^1,5
1Department of Radiology, Stanford University, Stanford, CA, United States, ²Department of Bioengineering, Stanford University, Stanford, CA, United States, ³GE HealthCare, Menlo Park, CA, United States, ⁴Department of Medicine, Stanford University, Stanford, CA, United States, ⁵Cardiovascular Institute, Stanford University, Stanford, CA, United States

Keywords: Flow, Machine Learning/Artificial Intelligence, Phase Contrast

We have previously trained a deep learning-based (DL) reconstruction for 2D PC-MRI using fully-sampled (n=194) raw k-space pediatric datasets. This DL-based reconstruction provided up to 9x undersampling with ≤5% error in accuracy and precision of peak velocity and total flow. Herein, we analyze this pediatric trained DL-based reconstruction in adult volunteers (n=3) and adult patients (n=8) and show that our DL-based reconstruction provides ~5% error in accuracy and precision of peak velocity and total flow for up to 7x undersampling.

View Presentation Video

Kelly Jarvis¹, Ethan Johnson¹, Haben Berhane¹, Adam Richter¹, Elizabeth Weiss¹, and Michael Markl^1
1Department of Radiology, Northwestern University, Chicago, IL, United States

Keywords: Flow, Data Processing

Analysis of pulse wave velocity by 4D flow MRI usually involves a manual processing pipeline with multiple steps of active use. We set out to automate the entire pipeline from original DICOMs to parameter quantification for evaluation of “whole-chest” 4D flow MRI. A deep learning algorithm trained on aortic acquisitions was used to facilitate segmentation. The supervised automation pipeline allowed for shorter analysis times and less user interaction, enabling the user to focus on refining aortic segmentations generated by deep learning. Results were comparable, but additional work including adjustment of preprocessing settings and retraining of CNN is warranted for optimization.

View Presentation Video

Aaron Christhoper Ponce^1,2, Sergio Uribe^2,3,4,5, and Julio Sotelo^2,5,6
1School of Informatics Engineering, Universidad de Valparaíso, Valparaíso, Chile, ²Millennium Institute for Intelligent Healthcare Engineering, iHEALTH, Santiago, Chile, ³Department of Radiology, School of Medicine, Pontificia Universidad Católica de Chile, 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, ⁶School of Biomedical Engineering, Universidad de Valparaíso, Valparaíso, Chile

Keywords: Flow, Segmentation

The segmentation of large vessels in 4D Flow MRI remains a challenge, due to different problems such as random acquisition noise, low spatial and temporal resolution, velocity aliasing, respiratory motion, phase offsets. For that reason the use of a single segmentation has been standardized to represent the geometry throughout all the cardiac phases. However, recent studies have proposed the use of image registration to be able to solve this problem. In this work, an algorithm based on a medical image registration neural network is proposed to improve the segmentation over time for the cardiac phases of the systole period. 

View Presentation Video

Stefano Buoso¹, Pietro Dirix¹, and Sebastian Kozerke^1
1Institute of Biomedical Engineering, ETH Zurich, Zurich, Switzerland

Keywords: Flow, Velocity & Flow

We propose physics-informed neural networks to augment 2D phase-contrast flow measurements in order to reconstruct full 3D velocity and pressure fields. In this conceptual work, 2D flow measurements are assimilated using the incompressible Navier-Stokes equations defined on the vessel anatomy reconstructed from anatomical scans. The network leverages a low-rank representation of velocity and pressure fields as physical prior and it is demonstrated to reconstruct 3D velocity and pressure gradient fields with good accuracy.

View Presentation Video

Huili Yang^1,2, Andrine Larsen³, Joshua D Robinson⁴, KyungPyo Hong¹, Cynthia K Rigsby⁵, and Daniel Kim^1,2
1Radiology, Northwestern University, Chicago, IL, United States, ²Biomedical Engineering, Northwestern University, Evanston, IL, United States, ³Biomedical Engineering, Lehigh University, Bethlehem, PA, United States, ⁴Pediatrics, Ann & Robert H. Lurie Children’s Hospital, Chicago, IL, United States, ⁵Medical Imaging, Ann & Robert H. Lurie Children’s Hospital, Chicago, IL, United States

Keywords: Flow, Cardiovascular, compressed sensing, real-time imaging, phase-contrast MRI

Highly-accelerated real-time phase-contrast (rt-PC) using compressed sensing may underestimate (~17%) velocity measurements due to over regularization. We sought to address this underestimation by incorporating view-sharing (VS) and k-space weighted image contrast (KWIC) filtering. This is particularly important for pediatric patients who have faster heart rates and smaller hearts than adults, it is particularly important to achieve high temporal resolution. The proposed reconstruction achieves 25 ms temporal resolution without significant temporal blurring artifacts.

View Presentation Video

Gloria Wolkerstorfer¹, Stefano Buoso¹, and Sebastian Kozerke^1
1Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland

Keywords: Flow, Data Analysis, Aorta; Shape modelling; Shape reconstruction; Image-based shape reconstruction; Aortic shape reconstruction

We propose an automatic pipeline to generate digital twin aortae from 2D+time MRI slices. First, a statistical shape model is fitted to the segmentation of 2D slices over time and then local refinements are applied to closely match dynamics over time. The approach is exemplified on a dataset of 10 aortic stenosis patients on which we quantify the impact of the number of available slices on the reconstructed anatomy.

View Presentation Video

Manuel A Morales¹, Amine Amyar¹, Siyeop Yoon¹, Jennifer Rodriguez¹, Martin S Maron², Ethan J Rowin², Shiro Nakamori¹, Jiwon Kim³, Robert M Judd⁴, Jonathan W Weinsaft³, Warren J Manning¹, and Reza Nezafat^1
1BIDMC, Boston, MA, United States, ²Tufts Medical Center, Boston, MA, United States, ³Weill Cornell Medicine, New York, NY, United States, ⁴Duke University, Durham, NC, United States

Keywords: Flow, Cardiovascular

Quantification of blood flow using 2D or 4D phase-contrast (PC) MRI is routinely being used to evaluate blood flow in cardiovascular disease. However, PC has only a modest temporal resolution compared to echocardiography. We sought to develop a deformation-encoding transformer (DENT) model for cardiac frame interpolation and evaluate its potential in increasing temporal resolution. DENT was trained using a large multi-center (centers = 3), multi-vendor (vendors = 3) and multi-field strength (1.5T, 3T) cine MRI dataset (patients = 3178). The model was successfully applied to 2D/4D-PC MRI without modifications, enabling a 2-fold gain in temporal resolution.

View Presentation Video

Vanessa Thiesen¹, Ansgar Adler¹, and Martin Uecker^1,2,3
1Institute of Biomedical Imaging, Graz, Austria, ²Institute for Diagnostic and Interventional Radiology of the University Center Göttingen, Göttingen, Germany, ³DZHK (German Center for Cardiovascular Research), Partner Site Göttingen, Göttingen, Germany

Keywords: Flow, Velocity & Flow

The assessment of flow within a heart cycle places high demands on the temporal resolution of the measurement protocol. With self-gating we can exploit the quasi-periodicity of cardiac and respiratory motion to meet this challenge. We combined a Hadamard flow-encoding scheme together with SSA-FARY to further elaborate on the idea of a flow measurement with a high temporal and spatial resolution. The approach delivered promising results in 2D and can be seen as a basis for further developement towards 3D flow MRI.

View Presentation Video

Jie Xiang¹, Jerome Lamy², Gigi Galiana^1,2, and Dana C. Peters^1,2
1Department of Biomedical Engineering, Yale University, New Haven, CT, United States, ²Department of Radiology and Biomedical Imaging, Yale University, New Haven, CT, United States

Keywords: Flow, Quantitative Imaging

Diastolic function evaluation requires estimates of early and late diastolic mitral valve flow velocities (E and A), normally evaluated by echocardiography, using phase contrast (PC). Our goal was to investigate k-t principal component analysis (PCA), constrained by either global and local principal components, for E and A measurements. Importantly, we found that retaining the central k-space used in estimating the principal components improved the final velocity maps. We compared local PCA with global PCA. Retrospectively undersampled images showed strong agreement with standard PC acquisitions. Higher E and A were observed using higher temporal resolution in prospective k-t PCA PC acquisitions.  

View Presentation Video

Guo Jiaxuan¹, Yue Xiuzheng², Huang shan², and Zhu Li^3
1Ningxia medical university, Yinchuan, China, ²Philips Healthcare, Beijing, China, ³General Hospital of Ningxia Medical University, Yinchuan, China

Keywords: Heart, Cardiovascular, Heart Failure

The number of patients with heart failure is increasing. Our study is based on the 4D flow technique of cardiac magnetic resonance and attempts to find cardiac functional parameters that precede structural changes in the heart at the level of the ventricular inflow and outflow tracts. We have found easier methods for visualizing blood flow and have the potential to further advance clinical diagnosis.

View Presentation Video

Tania Lala¹, Lea Christierson^2,3, Petter Frieberg¹, Daniel Giese^4,5, Nina Hakacova ², Pia Sjöberg¹, Ellen Ostenfeld¹, and Johannes Töger^1
1Clinical Physiology, Department of Clinical Sciences Lund, Skåne University Hospital, Lund University, Lund, Sweden, ²Department of Clinical Sciences Lund, Pediatric Heart Center, Skåne University Hospital, Lund University, Lund, Sweden, ³Department of Biomedical Engineering, Lund University, Lund, Sweden, ⁴Magnetic Resonance, Siemens Healthcare GmbH, Erlangen, Germany, ⁵Institute of Radiology, University Hospital Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany

Keywords: Flow, Velocity & Flow

Real-time Phase Contrast MRI with Compressed Sensing reconstruction is a promising method to image cardiac flow in patients with arrhythmia, where cardiac-gated sequences fail. In this study, we compared it against the gated standard clinical flow MRI using a phantom model and in patients. The phantom underwent pulsatile periodic and non-periodic flow. For in vivo validation, non-arrhythmic patient data (N=10) and data from patient with atrial fibrillation (N=1) were collected from the ascending aorta. Non-periodic flow was captured both in the phantom and in vivo. Real-time MRI showed good accuracy in net flow quantification and underestimated peak flow.

View Presentation Video

Naoya Kinota^1,2, Daisuke Abo³, Daisuke Kato¹, Satonori Tsuneta³, Noriko Nishioka¹, Kinya Ishizaka⁴, Noriyuki Fujima³, Kazuyuki Minowa⁵, and Kohsuke Kudo^1,6
1Department of Diagnostic Imaging, Faculty of Medicine and Graduate School of Medicine, Hokkaido University, Sapporo, Japan, ²Department of Dental Radiology, Hokkaido University Hospital, Sapporo, Japan, Japan, ³Department of Diagnostic and Interventional Radiology, Hokkaido University Hospital, Sapporo, Japan, ⁴Department of Radiological Technology, Hokkaido University Hospital, Sapporo, Japan, ⁵Department of Radiology, Faculty of Dental Medicine, Hokkaido University, Sapporo, Japan, ⁶Global Center for Biomedical Science and Engineering, Faculty of Medicine, Hokkaido University, Sapporo, Japan

Keywords: Vessels, Velocity & Flow

Echo-planer imaging-based three-dimensional cine phase contrast (4D flow) MRI has been reported to achieve good image quality with a short acquisition time in cardiovascular imaging, but its application to the portal venous (PV) system has not been reported. In this study, the turbo-field echo-planar imaging (TFEPI) 4D flow MRI with a turbo-field echo (TFE)-based one in the PV system was compared. The TFEPI sequence showed a better image quality and flow consistency at the splenic and superior mesenteric vein confluence with a shorter acquisition time compared to the conventional TFE-based one.