View Presentation Video

Andreas Kofler¹, Clemens Sirotenko², Felix Frederik Zimmermann¹, David Schote¹, Christoph Kolbitsch^1,3, Fatima Antarou Ba⁴, Fabian Altekrüger^4,5, Evangelos Papoutsellis^6,7, and Kostas Papafitsoros^8
1Physikalisch-Technische Bundesanstalt (PTB), Braunschweig and Berlin, Germany, ²Weierstraß-Institut für Angewandte Analysis und Stochastik, Berlin, Germany, ³King’s College London, London, United Kingdom, ⁴Technische Universität Berlin, Berlin, Germany, ⁵Humboldt Universität zu Berlin, Berlin, Germany, ⁶Science and Technology Facilities Council (STFC), Oxford, United Kingdom, ⁷Finden Ltd, Oxford, United Kingdom, ⁸Queen Mary University of London, London, United Kingdom

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, Interpretable Machine Learning, Algorithmic Unrolling, Iterative Neural Networks

We propose a method for estimating spatio-temporal regularization parameter-maps to be used for dynamic cardiac MR image reconstruction using total variation (TV)-minimization. Based on recent developments in algorithmic unrolling using Neural Networks (NNs), our approach uses two sub-networks. The first one predicts a spatio-temporal regularization parameter-map from an input image. Then, a second sub-network approximately solves a TV-reconstruction problem  which is formulated with the estimated regularization parameter-map. We show that the proposed method can be used to further improve the TV-reconstructions compared to using only one single scalar regularization parameter or two regularization parameters for space and time.

View Presentation Video

Guohao Jiang¹, Linzheng Hong¹, Zhaopeng Li², and Jun Xie^2
1ShanghaiTech University, Shanghai, China, ²United Imaging Healthcare, Shanghai, China

Keywords: Image Reconstruction, Diffusion/other diffusion imaging techniques

Fast liver DWI scan has significant clinical relevance, however, motion artifacts, long scan time and low SNR remain as major technical challenges for liver DWI. In this study, we proposed a joint recon method based on global and local high order tensor SVD (HOSVD), combining with shifted acquisition sampling pattern across diffusion directions. The proposed method improves DWI image quality and enables fast liver DWI scan within a single breath-hold.

View Presentation Video

Wahidul Alam¹, Junjie Liu², and Sajan Goud Lingala^1,3
1Roy J. Carver Department of Biomedical Engineering, University of Iowa, iowa city, IA, United States, ²Department of Neurology, University of Iowa, iowa city, IA, United States, ³Department of Radiology, University of Iowa, Iowa city, IA, United States

Keywords: Image Reconstruction, Head & Neck/ENT

Obstructive sleep apnea (OSA) is characterized by breathing-related obstructions of the upper airway during sleep. In this work, we develop a motion resolved extra-dimension sparsity-based approach to resolve the kinematics of upper-airway in OSA during natural sleep. This approach is demonstrated on two adult OSA patients: one undergoing a Bi-directional positive airway pressure therapy during MRI scanning, and one without therapy

View Presentation Video

Xuetong Zhou^1,2, Philip K. Lee², and Brian A. Hargreaves^1,2,3
1Department of Bioengineering, Stanford University, Stanford, CA, United States, ²Department of Radiology, Stanford University, Stanford, CA, United States, ³Department of Electrical Engineering, Stanford University, Stanford, CA, United States

Keywords: Image Reconstruction, Image Reconstruction

Water/fat separation is a reliable fat suppression technique. However, it is challenging in EPI-based imaging due to the large displacement along phase-encoding direction. We demonstrate the artifacts resulting from the poor-conditioning reconstruction of water/fat separation with EPI-based acquisition in body imaging. As a solution, unlike conventional multi-echo acquisition methods, our approach encodes the spectral components (water/fat) by acquiring and reconstructing the spatial and echo-time dimensions jointly. EPTI sampling trajectories and water/fat masking were used to improve the conditioning of the reconstruction. Phantom and in vivo (brain and breast) experiments were performed to show both results and limitations of this approach.

View Presentation Video

Thai Akasaka¹, Koji Fujimoto², Martijn Cloos³, Tomohisa Okada¹, Shinichi Urayama¹, and Isa Tadashi^1
1Human Brain Research Center, Kyoto University, Kyoto, Japan, ²Real World Data Research and Development, Kyoto University, Kyoto, Japan, ³University of Queensland, Brisbane, Australia

Keywords: Image Reconstruction, Image Reconstruction

The pulsation of intracranial aneurysms (IAs) is a novel risk factor of rupture. We aim to visualize the pulsation of IAs with 7T MRI using GRASP, a relatively novel MRI technique that provides high spatial and temporal resolution. Using thsis method, we were able to depict the pulsation of an aneurysm phantom as a 4D-cine movie. This technique may be useful in predicting the probability of IA rupture more accurately. 

View Presentation Video

Laura Pfaff^1,2, Fabian Wagner¹, Julian Hossbach^1,2, Elisabeth Preuhs¹, Fasil Gadjimuradov^1,2, Thomas Benkert², Dominik Nickel², Tobias Wuerfl², and Andreas Maier^1
1Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany, ²Magnetic Resonance, Siemens Healthcare GmbH, Erlangen, Germany

Keywords: Image Reconstruction, Diffusion/other diffusion imaging techniques

The diagnostic value of diffusion-weighted MR images is often degraded by their inherently low signal-to-noise ratio (SNR), especially for high b-values. In this context, the application of learning-based denoising methods is difficult since most methods require noise-free target images for training. We show how to denoise and evaluate diffusion-weighted MR images in a self-supervised manner by exploiting an adapted version of Stein’s unbiased risk estimator and specific properties of the data. Both quantitative and qualitative evaluations indicate increased performance over state-of-the-art unsupervised denoising methods.

View Presentation Video

Yasuhiro Goto¹, Michinobu Nagao², Masami Yoneyama³, Johannes M Peeters⁴, Yasutomo Katsumata³, Isao Shiina¹, Kazuo Kodaira¹, Yutaka Hamatani¹, Takumi Ogawa¹, Mana Kato¹, and Shuji Sakai^2
1Department of Radiological Services, Tokyo Women's Medical University, Tokyo, Japan, ²Department of Diagnostic imaging & Nuclear Medicine, Tokyo Women's Medical University, Tokyo, Japan, ³Philips Japan, Tokyo, Japan, ⁴Philips Healthcare, Best, Netherlands

Keywords: Image Reconstruction, Blood vessels

In this study demonstrated that the higher reduction factor (R=10.0) with a one-minute scan still provided sufficient image quality with significantly faster scan time compared with conventional REPI-SENSE. REPIX would be useful for further assessment of PAD pathology even with multiple VENC acquisitions still in a clinically feasible scan time.REPIX 4D-MRA well depicted peripheral arteries clearly in a significantly short time thanks to Compressed SENSE, compared with conventional SENSE.

View Presentation Video

Andrew Phair¹, Anastasia Fotaki¹, Lina Felsner¹, Haikun Qi², René M. Botnar¹, and Claudia Prieto^1
1School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, ²School of Biomedical Engineering, ShanghaiTech University, Shanghai, China

Keywords: Image Reconstruction, Cardiovascular

A deep learning reconstruction framework, trained in an end-to-end fashion and incorporating both a non-rigid respiratory motion estimation network and a motion-informed model-based reconstruction network, has been previously demonstrated to enable good quality images from seven-fold undersampled acquisitions for coronary magnetic resonance angiography applications. Herein, we apply the framework to whole-heart MRI scans of patients with congenital heart disease, enabling fast reconstruction of 7×-accelerated acquisitions and achieving image quality comparable to that of state-of-the-art patch-based low-rank iterative techniques.

View Presentation Video

Faisal Najeeb¹, Madiha Arshad ¹, Muhammad Shafique¹, and Hammad Omer^1
1Electrical and Computer Engineering, COMSATS University Islamabad, Islamabad, Pakistan

Keywords: Image Reconstruction, Artifacts, Deep Learning, Compressed Sensing , Parallel MRI

Most deep learning methods apply U-Net either in image or k-space domain. Nevertheless, these methods have limitations: (1) Directly applying U-Net in k-space domain is not optimal for extracting features; (2) conventional image-domain oriented U-Net does not fully utilize the information of encoder part of the network for extracting features in the decoder part. In this paper, a dual-domain deep learning-based approach is presented, incorporating multi-coil data consistency layers for the reconstruction of cardiac MR images from 1-D Variable Density (VD) under-sampled data. Experiments show superior reconstruction results of the proposed method than conventional Compressed Sensing (CS) method.

View Presentation Video

Jörn Huber¹, Laurent Ruck^2,3, Matthias Günther^1,4,5, Armin Nagel^2,6, and Simon Konstandin^1,4
1Fraunhofer Institute for Digital Medicine MEVIS, Bremen, Germany, ²Institute of Radiology, University Hospital Erlangen, Erlangen, Germany, ³Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Erlangen, Germany, Erlangen, Germany, ⁴mediri GmbH, Heidelberg, Germany, ⁵Faculty 1 (Physics/Electrical Engineering), University of Bremen, Bremen, Germany, ⁶Division of Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany, Erlangen, Germany

Keywords: Image Reconstruction, Myocardium

Motion correction in interleaved 23Na/1H MRI of the human heart is important to improve the diagnostic reliability of reconstructed images and derived quantitative parameters. Therefore, this work demonstrates and compares the application of different reconstruction techniques to undersampled and motion-gated 23Na/1H MRI data at 7 T.

View Presentation Video

Wolfgang G Rehwald^1,2, Kelvin Chow³, Rafael Rojas⁴, Nestor Mena⁴, Diana Alexandrov⁴, George Gamoneda⁴, David Wendell⁴, Ryan Seward⁴, Jeana Dement⁴, Vera Kimbrell⁴, Han Kim⁴, Indraneel Borgohain⁵, Igor Klem⁴, and Raymond Kim^4
1Siemens Healthineers, Durham, NC, United States, ²Duke Cardiovascular MR Center, Duke University, Durham, NC, United States, ³Siemens Healthineers, Chicago, IL, United States, ⁴Duke Cardiovascular MR Center, Duke University Hospital, Durham, NC, United States, ⁵Siemens Healthineers, Princeton, NJ, United States

Keywords: Image Reconstruction, Motion Correction, Reordering

We introduce a free breathing segmented LGE technique combining a new acquisition, reordering and reconstruction scheme. The on-scanner reconstruction uses MATLAB and FIRE. The method produces image quality (IQ) and SNR otherwise only obtainable with breath held segmented interleaved LGE. In 27 patients, we show that IQ is superior to free breathing segmented interleaved LGE with multiple averages. SNR is higher compared to averaged motion corrected single shots when matching spatial and temporal resolution, the number of used measurements per image, and the readout type. Being a segmented technique, temporal and spatial resolution limitations of single shots do not apply.

 

View Presentation Video

Yilin Liu¹, Asala Ahamd¹, Xingyu Nie², and Guang (George) Li^1
1Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, United States, ²Radiology, University of Kentucky, Lexington, KY, United States

Keywords: Image Reconstruction, Radiotherapy, time-resolved 4DMRI, time-resolved 4DCT, multi-breathing cycles

This study has demonstrated that the feasibility to reconstruct multiple-breath TR-4DCT via the super-resolution reconstruction framework through either CT^4D→(MR^BH→MR^FB) or (CT^4D←MR^BH)→ MR^FB deformable image registration. Using TR-4DCT, potential dosimetry consequences in radiotherapy of lung, liver, and pancreatic patients due to patient breathing irregularities can be readily assessed.

View Presentation Video

Andreas Kofler¹, Christian Wald², Tobias Schaeffter^1,3,4, Markus Haltmeier⁵, and Christoph Kolbitsch^1,3
1Physikalisch-Technische Bundesanstalt (PTB), Braunschweig and Berlin, Germany, ²Charité - Universitätsmedizin Berlin, Berlin, Germany, ³King’s College London, London, United Kingdom, ⁴Technical University of Berlin, Berlin, Germany, ⁵University of Innsbruck, Innsbruck, Austria

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, Signal Processing, Sparsity-based Methods, Convolutional Dictionary Learning

In this work we consider three different variants physics-informed Neural Networks (PINNs) which use the convolutional dictionary learning framework for image reconsruction in dynamic cardiac MRI. Although all three NNs share the same mechanism for regularization, the iterative schemes differ because they are derived from different problem formulations. We compare the  methdos in terms of reconstruction performance as well as stability with respect to the number of iterations. All three methods yield similarly accurate reconstructions. However, by  construction, only one of the three methods defines a convergent reconstruction algorithm and is therefore stable w.r.t. to the number of iterations.

View Presentation Video

Aaron Curtis^1,2 and Hai-Ling Margaret Cheng^1,2,3
1Electrical and Computer Engineering, University of Toronto, Toronto, ON, Canada, ²Translational Biology & Engineering Program, Ted Rogers Centre for Heart Research, Toronto, ON, Canada, ³Institute of Biomedical Engineering, University of Toronto, Toronto, ON, Canada

Keywords: Image Reconstruction, Heart

Robust, real-time dynamic cardiac MRI (CMR) would provide information on the temporal signatures of disease that we currently cannot assess. We present a novel Kalman filtering framework that uses a priori statistics derived from a single cardiac cycle to adaptively predict temporal cardiac dynamics. Kalman filtering is ideal, as it ameliorates noise introduced from our maximum acceleration factor of 60, guarantees reconstruction fidelity, and enables flexible undersampling. Furthermore, reconstruction may be performed at an even higher temporal resolution than the training data. As such, our algorithm can be a foundation for true real-time dynamic CMR.

Junhao Zhang^1,2, Yujiao Zhao^1,2, Jiahao Hu^1,2,3, Ye Ding^1,2, Christopher Men^1,2, Vick Lau^1,2, Alex T.L.Leong^1,2, and Ed X. Wu^1,2
1Laboratory of Biomedical Imaging and Signal Processing, the University of Hong Kong, HongKong, China, ²Department of Electrical and Electronic Engineering, the University of Hong Kong, HongKong, China, ³Department of Electrical and Electronic Engineering, Southern University of Science and Technology, Shenzhen, China

Keywords: Data Analysis, Cardiovascular, cardiac reconstruction,deep learning

We present a CNN-based deep learning model to reconstruct the cardiac cine images from undersampled single-channel 4D MR data.  The wavelet transform and spatial-temporal attention mechanisms are introduced in the model. The proposed model could reconstruct the cardiac images and recover the cardiac wall motion more robustly than the low-rank plus sparsity (i.e., L+S) reconstruction method. This approach presents one promising solution for accelerated cardiac dynamic imaging with one single channel through deep learning from the sparsity in spatial and temporal domains.

View Presentation Video

Xiaokai Wang¹, Jiayue Cao², Kuan Han², Minkyu Choi², Yushi She², Ulrich Scheven², and Zhongming Liu^2
1Biomedical Engineering, University of Michgan, Ann Arbor, MI, United States, ²University of Michigan, Ann Arbor, MI, United States

Keywords: Data Analysis, Data Analysis, Neural Network

We describe a method, namely neural ordinary differential equations, to track the movement of the stomach based on dynamic and contrast-enhanced gastrointestinal MRI. This model uses a neural network to learn the continuous biomechanical process that drives the shape change of the stomach wall over the course of digestion. This method allows us to represent gastric motor events on a generic surface template of the stomach and to further reveal the pattern of gastric motility with higher specificity and resolution than are previously attainable in vivo. This method should be also applicable to other organs, such as the heart.

View Presentation Video

Can Wu¹, Guang (George) Li¹, and Yilin Liu^1
1Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, United States

Keywords: Image Reconstruction, Radiotherapy, TR-4DMRI, multi-breathing cycles, breathing irregularities, image quality, compressed sensing

Since the development of respiratory-correlated four-dimensional computed tomography (RC-4DCT) [1] and magnetic resonance imaging (RC-4DMRI) [2,3], patient-specific respiratory-induced tumor motion has been incorporated in radiotherapy for treating mobile tumors, such as lung, liver and pancreatic cancer.   However, the RC-based snapshot 4D imaging only provides one breathing cycle, often contains severe binning motion artifacts, and may not represent tumor motion over 20-minute treatment, affecting treatment outcomes.  Therefore, respiratory motion irregularities remain a challenge in radiotherapy.  In this study, we report an improved time-resolved 4DMRI technique that captures multi-breath and can be used clinically and quantifies tumor motion irregularities.

View Presentation Video

Xin Li¹, Travis Rice-Stitt², Lina Gao³, Marina Aguiñaga⁴, Kevin R Turner⁵, Bryan Foster⁶, Fergus Coakley⁶, Mark Garzotto^4,7, and Ryan Kopp^4,7
1Advanced Imaging Research Center, Oregon Health & Science University, Portland, OR, United States, ²Pathology, Oregon Health & Science University, Portland, OR, United States, ³Knight Biostatistics Shared Resources, Oregon Health & Science University, Portland, OR, United States, ⁴Urology, Oregon Health & Science University, Portland, OR, United States, ⁵Providence Health and Service, Portland, OR, United States, ⁶Diagnostic Radiology, Oregon Health & Science University, Portland, OR, United States, ⁷Portland VA Medical Center, Portland, OR, United States

Keywords: Data Analysis, DSC & DCE Perfusion

Recent advance in data acquisition makes fast DCE-MRI data acquisition feasible.  This work investigated the impact of DCE-MRI temporal resolution on the data’s pharmacokinetic modeling and the water exchange effect imparted into the model parameters. Using both the Tofts model and the water-exchange sensitized Shutter-Speed model, our results showed that DCE data with higher temporal resolution (shorter intersample interval) is beneficial in model parameter precision and in quantifying transcytolemmal water exchange effect. The later may offer additional lesion-detection specificity in clinical prostate MRI. 

View Presentation Video

Samuel Perron¹, Matthew S. Fox^1,2, and Alexei V. Ouriadov^1,2,3
1Physics and Astronomy, University of Western Ontario, London, ON, Canada, ²Lawson Health Research Institute, London, ON, Canada, ³School of Biomedical Engineering, University of Western Ontario, London, ON, Canada

Keywords: Data Acquisition, Low-Field MRI

Accelerated MRI could significantly improve image quality of lung imaging without increasing costs, especially for low field strengths. The signal decay of a series of undersampled xenon-129 images is fitted to the Stretched-Exponential-Model to yield higher SNR. The proposed method was implemented in undersampled phantom images at low field (0.074T) and human lung images at high field (3T) using the FGRE pulse sequence with hyperpolarized ¹²⁹Xe; SNR was significantly improved within the same scan duration compared to a fully-sampled image. Potential issues with the compressed-sensing reconstruction are identified and possible solutions presented to promote robustness of method.

View Presentation Video

Robin Ferincz¹, Ludovica Romanin¹, Jérôme Yerly², Davide Piccini^1,3,4, Matthias Stuber², and Christopher William Roy^1
1Department of Radiology, Lausanne University Hospital (CHUV) and University of Lausanne (UNIL), Lausanne, Switzerland, ²Center for Biomedical Imaging (CIBM), Lausanne, Switzerland, ³Advanced Clinical Imaging Technology (ACIT), Siemens Healthineers International AG, Lausanne, Switzerland, ⁴LTS5, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland

Keywords: Data Acquisition, Artifacts

Recent advances have enabled high resolution cardiac and respiratory motion-resolved 5D whole-heart MRI using non-contrast enhanced non-selective 3D radial bSSFP. However, the nature of bSSFP, the subject dependent anatomy, as well as the underlying sparse reconstruction can lead to banding and streaking artifacts, which degrade image quality and reduce diagnostic utility. In this work, the impact of slab-selective RF pulses in a programmable orientation that is independent of the k-space trajectory is assessed in a cohort of healthy volunteers. Preliminary results suggest that a subject-specific slab orientation can reduce artifacts and improve image quality.