ISSN# 1545-4428 | Published date: 19 April, 2024
You must be logged in to view entire program, abstracts, and syllabi
At-A-Glance Session Detail
   
Quantitative Image Reconstruction
Oral
Acquisition & Reconstruction
Tuesday, 07 May 2024
Hall 606
15:45 -  17:45
Moderators: Hongyan Liu & Tobias Kober
Session Number: O-02
CME Credit

15:450622.
Simultaneous 3D T1, T2, and fat-fraction mapping with respiratory-motion correction, for comprehensive liver tissue characterisation at 0.55T
Donovan Tripp1, Radhouene Neji1, Karl P Kunze1,2, Michael G Crabb1, Claudia Prieto1,3,4, and René Botnar1,3,4,5
1School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, 2MR Research Collaborations, Siemens Healthcare Limited, Camberley, United Kingdom, 3School of Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile, 4Millennium Institute for Intelligent Healthcare Engineering, Santiago, Chile, 5Institute for Biological and Medical Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile

Keywords: Quantitative Imaging, Quantitative Imaging, Liver, Low-Field

Motivation: Mulitparametric quantitative MRI is a powerful tool for diagnosis of liver disease, but current clinical sequences will acquire 2D slices in separate scans, prone to misregistration.

Goal(s): Demonstrate the simultaneous in-vivo acquisition of T1, T2, and fat fraction maps over the whole liver from a single free-breathing scan at 0.55T.

Approach: A dictionary-matching-based framework with non-rigid respiratory motion corrected reconstruction was validated in a cohort of ten healthy subjects.

Results: T1, T2, and fat fraction values acquired in phantoms and in vivo showed good agreement with values from corresponding reference scans.

Impact: Our technique promises an efficient means to acquire multiple parameter maps providing comprehensive staging and diagnosis of non-alcoholic fatty liver disease, believed to affect over two billion people worldwide.

15:570623.
Accelerating Quantitative MRI using Self-supervised Deep Learning with Model Reinforcement
Wanyu Bian1,2, Albert Jang1,2, and Fang Liu1,2
1Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, 2Harvard Medical School, Boston, MA, United States

Keywords: Quantitative Imaging, Quantitative Imaging, Model-based Reconstruction, Relaxometry, Brain, Self-supervised Learning

Motivation: Quantitative MRI (qMRI) is time-consuming and requires substantial efforts for acceleration to cut down the acquisition time.

Goal(s): This paper proposes a novel self-supervised learning framework that uses model reinforcement, RELAX-MORE, for accelerated qMRI reconstruction.

Approach: The proposed method uses an optimization algorithm to unroll an iterative model-based qMRI reconstruction into a deep learning framework, enabling accelerated MR parameter maps that are highly accurate and robust.

Results: The proposed method generates high quality MR parameter maps that correct for image artifacts, removes noise, and recovers image features in regions of imperfect image conditions.

Impact: This work demonstrates the feasibility of a new self-supervised learning method for rapid MR parameter mapping, that is readily adaptable to the clinical translation of qMRI.

16:090624.
MRI2Qmap: compressed-sampled multiparametric quantitative MRI reconstruction using learned spatial priors from multimodal MRI datasets
Mohammad Golbabaee1, Matteo Cencini2, Carolin M Pirkl3, Marion I Menzel3, Michela Tosetti4, and Bjoern H Menze5
1University of Bristol, Bristol, United Kingdom, 2INFN Pisa division, Pisa, Italy, 3GE Healthcare, Munich, Germany, 4IRCCS Stella Maris, Pisa, Italy, 5University of Zurich, Zurich, Switzerland

Keywords: MR Fingerprinting, Quantitative Imaging, MR Fingerprinting, Compressed sensing, Image reconstruction, AI/ML Image Reconstruction

Motivation: Deep learning excels at compressed-sensing image reconstruction given large training datasets. Applying this paradigm to accelerated quantitative MRI, including magnetic resonance fingerprinting (MRF), is challenging because quantitative imaging datasets for training are scarce.

Goal(s): Can we overcome this limitation using new sources of training data from routine, largely available weighted-MRI images?

Approach: We introduce MRI2Qmap, a plug-and-play quantitative image reconstruction algorithm based on deep image denoising models pretrained on large multimodal weighted-MRI datasets.

Results: We showed, for the first time, that spatial/structural priors learned from independently-acquired datasets of routine weighted-MRI images can be effectively used for quantitative MRI image reconstruction.

Impact: Thanks to the widespread use of MRIs, our approach could enable much larger datasets to be used for training potentially enhanced AI models for fast quantitative MRI/MRF image reconstruction.

16:210625.
Rapid Pediatric Imaging with Zero-Shot Deep Subspace Reconstruction for Multiparametric Quantitative MRI
Yohan Jun1,2, Shohei Fujita1,2, Jaejin Cho1,2,3, Xingwang Yong1,2,4, Eugene Milshteyn5, Camilo Jaimes2,3,6, Suely Fazio Ferraciolli2,3,6, Borjan Gagoski2,7, Michael S Gee2,3,6, and Berkin Bilgic1,2,8
1Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston, MA, United States, 2Department of Radiology, Harvard Medical School, Boston, MA, United States, 3Pediatric Imaging Research Center, Massachusetts General Hospital, Boston, MA, United States, 4Zhejiang University, Hangzhou, China, 5GE Healthcare, Boston, MA, United States, 6Department of Radiology, Massachusetts General Hospital, Boston, MA, United States, 7Fetal-Neonatal Neuroimaging & Developmental Science Center, Boston Children’s Hospital, Boston, MA, United States, 8Harvard/MIT Health Sciences and Technology, Cambridge, MA, United States

Keywords: Quantitative Imaging, Pediatric, Quantitative Imaging

Motivation: To address unmet needs for accurate, rapid, and high-fidelity quantitative MRI using a 3D-QALAS sequence.

Goal(s): To enable accurate T1 and T2 mapping with reduced biases, g-factor noise amplification, and relaxation-related blurring compared to conventional QALAS.

Approach: We employed a zero-shot self-supervised subspace reconstruction technique, Zero-DeepSub, which combines scan-specific deep-learning-based reconstruction with low-rank subspace modeling, and demonstrated the performance using ISMRM/NIST phantom and pediatric patients.

Results: Zero-DeepSub enabled a highly accelerated, 2 min acquisition at 1 mm isotropic resolution at 3T, as well as a 5 min pediatric exam at 1.2 mm isotropic resolution at 1.5T.

Impact: Zero-DeepSub enabled accurate T1 and T2 mapping with reduced biases, g-factor noise amplification, and relaxation-related blurring, showing the potential to substantially speed up pediatric brain exams, thus obviating the need for or reducing the amount of sedation and anesthesia.

16:330626.
Sub-Second GRASP-LLR DCE: Locally Low-Rank Subspace Constraint aided by Deep Learning
Eddy Solomon1,2, Jonghyun Bae1, Linda Moy2, Laura Heacock2, Li Feng2, and Sungheon Gene Kim1,2
1Radiology, Weill Cornell Medicine, New York, NY, United States, 2Center for Advanced Imaging Innovation and Research (CAI2R), Department of Radiology, New York University, New York, NY, United States

Keywords: Quantitative Imaging, Breast, DCE

Motivation: We hope to advance the assessment of breast dynamic contrast-enhanced MRI (DCE-MRI) by enhancing image quality, temporal resolution, and temporal fidelity.

Goal(s): Propose a new radial GRASP reconstruction pipeline for DCE-MRI, which enables reliable spatially localized dynamics at a sub-second temporal resolution.

Approach: Presenting globally and locally low-rank reconstruction approaches for GRASP DCE-MRI aided by Residual Network (ResNet) architecture.

Results: Our results suggest that GRASP-LLR offers not only enhanced tumor lesion delineation with reduced background noise but also good separation between healthy, benign, and malignant cases.

Impact: We propose a new radial reconstruction pipeline for DCE-MRI which leverages a locally low-rank (LLR) subspace model in combination with deep learning approach, resulting in reliable spatially localized dynamics at a sub-second temporal resolution.

16:450627.
Modeling Phase Errors for Robust and Efficient Multidimensional MR Fingerprinting for Simultaneous Relaxation and Diffusion Mapping
Zhilang Qiu1, Siyuan Hu1, Walter Zhao1, Ken Sakaie2, Filip Szczepankiewicz3, Jessie E.P. Sun4, Mark A. Griswold4, Derek K. Jones5, and Dan Ma1
1Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States, 2Imaging Institute, Cleveland Clinic, Cleveland, OH, United States, 3Medical Radiation Physics, Clinical Sciences Lund, Lund University, Lund, Sweden, 4Department of Radiology, Case Western Reserve University, Cleveland, OH, United States, 5Cardiff University Brain Research Imaging Centre (CUBRIC), School of Psychology, Cardiff University, Cardiff, United Kingdom

Keywords: MR Fingerprinting, MR Fingerprinting

Motivation: Diffusion MRI can be corrupted by phase errors due to physiological motion, bulk motion, eddy currents, and other system imperfections, which makes its efficient embedding into MR Fingerprinting challenging.

Goal(s): To develop a new approach to correct artifacts in multidimensional MR Fingerprinting (mdMRF) for simultaneous relaxation and diffusion quantification, that obviates cardiac gating, motion compensation, navigators, or data removal.

Approach: Modeling potential phase errors using phase offset and phase dispersion during dictionary generation, then quantifying and correcting measured phase errors in dictionary matching.

Results: The proposed approach significantly mitigates artifacts in mdMRF diffusion parameter mapping.

Impact: Phase error-induced artifacts due to physiological motion, bulk motion, and eddy currents is a key limitation in diffusion MRI. We develop an approach to improve robustness and efficiency of artifact correction in multidimensional MR Fingerprinting for relaxation and diffusion mapping.

16:570628.
Towards ultrafast submillimeter T2* and QSM quantification at 3T using spherical Echo Planar Time Resolved Imaging (sEPTI)
Nan Wang1, Mark Nishimura2, Mahmut Yurt2, Mengze Gao1, Daniel Abraham2, Cagan Alkan2, Congyu Liao1, Xiaozhi Cao1, Zihan Zhou1, and Kawin Setsompop1
1Radiology Department, Stanford University, Stanford, CA, United States, 2Electrical Engineering, Stanford University, Stanford, CA, United States

Keywords: Image Reconstruction, Image Reconstruction

Motivation: To further improve the image quality, SNR, and reduce scan time by reducing averages for submillimeter T2* and QSM at 3T 

Goal(s): To achieve whole-brain 0.75-mm T2* and QSM quantification using sEPTI within a single average scan

Approach: We developed: (1) an iterative data-driven B0 update pipeline for accurate and high SNR B0 map; (2) data-driven eddy-current correction approach to reduce artifacts; (3) a physics-informed unrolled network to boost the SNR of the reconstructed image to achieve 2X acceleration by reducing the need of averages.

Results: sEPTI achieved whole-brain 0.75-mm T2* and QSM quantification within 84 seconds with the potential for wide applications

Impact: The work presented synergetic improvements in B0 update, eddy-current correction, and unrolled-network based SNR-boosted reconstruction for sEPTI, which achieves whole-brain 0.75-mm distortion-free and blurring-free T2* and QSM quantification at 3T in 84 seconds with the potentials for wide applications.

17:090629.
Time-Resolved Cardiac function: Myocardium Strain and First-Pass Perfusion Using MR-MOTUS
Thomas E. Olausson1,2, Maarten L. Terpstra1,2, Niek R.F. Huttinga1,2, Casper Beijst2, Niels Blanken3, Dominika Suchá3, Teresa Correia4,5, Cornelis A.T. van den Berg1,2, and Alessandro Sbrizzi1,2
1Computational Imaging Group for MR Therapy and Diagnostics, University Medical Center Utrecht, Utrecht, Netherlands, 2UMC Utrecht Cancer Center, Department of Radiotherapy, University Medical Center Utrecht, University Medical Center Utrecht, Utrecht, Netherlands, 3Department of Radiology, University Medical Centre Utrecht, University Medical Center Utrecht, Utrecht, Netherlands, 4School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, 5Centre for Marine Sciences (CCMAR), Faro, Portugal

Keywords: Myocardium, Image Reconstruction, First-Pass Myocardial Perfusion; Motion estimation; Motion correction; Low-Rank & Sparse; Time-resolved imaging; Cine; Dynamic; Perfusion; Cardiac

Motivation: We aim at improving accuracy and reliability in cardiac imaging for coronary artery disease diagnosis and management. We address motion artifacts in first-pass myocardial perfusion MR imaging.

Goal(s): To develop and validate motion correction techniques for motion field accuracy and strain quantification in non-ECG triggered myocardial first pass perfusion examinations.

Approach: We use a modified MR-MOTUS framework for motion separation and reconstruction in patient data.

Results: Our approach demonstrates higher accuracy in respiratory/cardiac motion field estimation with additional strain analysis.

Impact: Our findings have the potential to improve patient care by enabling free-breathing and non-ECG triggered examinations of myocardial first-pass perfusion simultaneously with strain analysis. We also open avenues for further research in cardiac imaging and motion correction techniques.

17:210630.
Joint optimization of multi-echo reconstruction and quantitative map estimation in Looping Star
Haowei Xiang1, Ilhan Kemal Onder2, Anahita H Mehta2, Jeffrey A Fessler1, and Douglas C Noll3
1Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI, United States, 2Department of Otolaryngology-Head and Neck Surgery, University of Michigan, Ann Arbor, MI, United States, 3Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States

Keywords: Image Reconstruction, Quantitative Imaging, Reconstruction

Motivation: Looping-star sequences, despite their advantages, exhibit low SNR and undersampling artifacts compared to standard GRE sequences.

Goal(s): This work proposes to jointly reconstruct multi-echo data and estimate quantitative maps in looping-star to boost the SNR, reduce the undersampling artifacts, and improve image quality.

Approach: Our approach frames echo image reconstruction and quantitative map estimation as a unified optimization problem. This is then split into two sub-problems, addressed alternately using CG-SENSE.

Results: Compared to individual echo reconstruction, our joint optimization improves tSNR of both echo images and T2* maps and effectively mitigates image artifacts.

Impact:  Our method jointly reconstructs multi-echo data in looping-star, enhancing SNR and reducing artifacts, with a notable tSNR improvement. It can be adapted to the Looping-Star fMRI protocol to potentially improve functional activity estimation.

17:330631.
Temporal Structured Low-Rank Reconstruction for First-Pass Myocardial Perfusion Imaging
Xi Chen1,2, Debiao Li1,3, and Anthony G. Christodoulou1,2
1Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, Los Angeles, CA, United States, 2Department of Radiological Sciences, David Geffen School of Medicine at UCLA, Los Angeles, CA, United States, 3Department of Bioengineering, UCLA, Los Angeles, CA, United States

Keywords: Sparse & Low-Rank Models, Perfusion, First-pass myocardial perfusion; structured low-rank

Motivation: First-pass myocardial perfusion imaging is a powerful tool for assessing coronary artery disease, but needs high levels of undersampling to achieve sufficient spatial coverage, spatiotemporal resolution, and motion robustness.

Goal(s): To develop efficient temporal image reconstruction models which can leverage linear time-invariant models of dynamic contrast enhancement without identifying an arterial input function or assuming tissue transfer function shapes.

Approach: We propose a novel temporal structured low-rank modeling technique to implicitly leverage linear time-invariant models of dynamic contrast enhancement.

Results: Temporal structured low-rank modeling outperforms conventional low-rank methods, especially as a local constraint.

Impact: Temporal structured low-rank modeling has the potential to improve spatial coverage, spatial resolution, and/or motion robustness for first-pass myocardial perfusion MRI.