08:15 | 0037.
| Four-dimensional iterative motion correction (iMoCO) for isotropic stack-of-spirals cine imaging at 0.55T Rajiv Ramasawmy1, Ahsan Javed1, Daniel Herzka2, Prakash Kumar3, Krishna Nayak3, Robert Lederman1, and Adrienne Campbell-Washburn1 1National Heart, Lung and Blood Institute, Bethesda, MD, United States, 2Case Western Reserve University and University Hospitals, Cleveland, OH, United States, 3University of Southern California, Los Angeles, CA, United States Keywords: Image Reconstruction, Cardiovascular, Image Reconstruction, Low-Field MRI, Data Acquisition Motivation: Three-dimensional (3D) isotropic cine imaging can be resliced and resampled for clinical diagnosis and planning for structural interventions. Currently, these 3D cine approaches are hampered by long scan times. Goal(s): To demonstrate cardiac-resolved iterative motion compensation (iMoCo) for a free-breathing stack-of-spirals 3D cine with optimized acquisition ordering at 0.55T Approach: The 3D cine was acquired in five healthy volunteers and one patient, reconstructed with cardiac-resolved iMoCo, and compared to a reference 2D cine. Results: The proposed method yielded high quality 3D cines. Volumetric measurements had good agreement with reference data (-2.3 ± 2.8% and 3.9 ± 10.4% in diastole and systole respectively). Impact: A gaussian-distributed stack-of-spirals sampling scheme paired with an iMoCo reconstruction improves image quality and sharpness for isotropic three-dimensional cines. This technique can be a useful tool for interventional planning and assessment or as a one-stop shop for diagnostic cardiovascular MRI. |
08:27 | 0038.
| Multi-scale plug-and-play energy framework for inverse problems Jyothi Rikhab Chand1 and Mathews Jacob1 1University of Iowa, Iowa city, IA, United States Keywords: Image Reconstruction, Data Processing Motivation: Unrolled algorithms provide high quality image reconstruction. However, their training is memory-intensive and is sensitive to forward model mismatches. Goal(s): To develop a memory-efficient plug-and-play algorithm, whose performance is comparable to unrolled algorithms and can be used with arbitrary forward models. Approach: We propose a memory-efficient energy-based multi-scale framework. We model the negative log prior with different smoothnesses using Convolutional Neural Networks (CNN). This approach enables us to relax the constraints on the CNN, while the multi-scale strategy improves the convergence to the global minimum. Results: The enhancements improves performance, making it comparable to end-to-end methods, while being robust to model mismatch. Impact: The proposed framework is memory-efficient compared to unrolled algorithms, paving the way for its usage in large-dimensional inverse problems. Its flexibility enables recovery of images with arbitrary forward operators.
|
08:39 | 0039.
| Time-Resolved Biomechanics using Spectro-Dynamic MRI: Proof of Principle in the Muscles of the Thigh Max H.C. van Riel1, David G.J. Heesterbeek1, Martijn Froeling1, Tristan van Leeuwen2, Cornelis A.T. van den Berg1, and Alessandro Sbrizzi1 1Department of Radiotherapy, Computational Imaging Group for MR Diagnostics and Therapy, UMC Utrecht, Utrecht, Netherlands, 2Mathematical Institute, Utrecht University, Utrecht, Netherlands Keywords: Image Reconstruction, Muscle, Time-Resolved, Motion, Strain Motivation: Measurements of biomechanical tissue properties require time-resolved reconstructions from dynamic experiments. The Spectro-Dynamic MRI framework achieves this by working directly from k-space data. Goal(s): To develop an experimental setup and reconstruction method with which time-resolved biomechanical information can be measured in vivo. Approach: An inflatable pressure cuff deformed the thigh muscles of a volunteer. Time-resolved images and strain maps were reconstructed directly from k-space data using the Spectro-Dynamic MRI framework. Results: Principal strains were obtained for different muscles in the thigh at a temporal resolution of 352 ms. The first principal strain direction could differentiate between muscle structures, indicating different underlying biomechanical properties. Impact: The reconstruction of time-resolved images and strains using Spectro-Dynamic MRI allows for time-resolved measurements of biomechanical parameters during dynamic loads with a straightforward experimental setup. This information is useful for studying the mechanical behavior of tissues. |
08:51 | 0040.
| DeepGrasp4D: A General Framework for Highly-Accelerated Real-Time 4D Golden-Angle Radial MRI Using Deep Learning Haoyang Pei1,2,3, Hersh Chandarana1,2, Daniel K Sodickson1,2, and Li Feng1,2 1Bernard and Irene Schwartz Center for Biomedical Imaging, Department of Radiology, New York University Grossman School of Medicine, New York City, NY, United States, 2Center for Advanced Imaging Innovation and Research (CAI2R), Department of Radiology, New York University Grossman School of Medicine, New York City, NY, United States, 3Department of Electrical and Computer Engineering, NYU Tandon School of Engineering,, New York City, NY, United States Keywords: Image Reconstruction, Image Reconstruction Motivation: Time-resolved real-time 4D MRI demands high imaging speed to achieve high spatial and temporal resolution. While conventional iterative reconstruction methods can accomplish this, they require substantial temporal correlations and impose a significant computational burden. Goal(s): This study proposes DeepGrasp4D, a deep learning technique tailored to efficiently reconstruct real-time 4D MR images with reduced temporal correlations and shortened scan times. Approach: DeepGrasp4D was developed based on an unrolled network that incorporates an explicit low-rank constraint and a temporal total variation constraint, enabling efficient reconstruction of 4D images from continuously acquired golden-angle radial k-space. Results: DeepGrasp4D enables accurate 4D MRI reconstruction at high acceleration rates. Impact: The proposed DeepGrasp4D technique enables efficient and reliable 4D MRI reconstruction from golden-angle radial data acquired with shortened scan times and reduced temporal correlations. This can be useful in various applications such as DCE-MRI or MRI-guided radiotherapy. |
09:03 | 0041.
| Image reconstruction for an 8-element loop-dipole rotating RF coil array (RRFCA) using a novel calibration-free GRAPPA-based method Lachlan West1, Andrew Phair2,3, Mingyan Li1, Michael Brideson3, Andrew P Bassom3, and Feng Liu1 1University of Queensland, Brisbane, Australia, 2King's College London, London, United Kingdom, 3University of Tasmania, Hobart, Australia Keywords: Image Reconstruction, Image Reconstruction Motivation: SENSE-based reconstruction is challenging for clinical imaging when rotating the RRFCA into multiple positions; therefore, a novel calibration-free GRAPPA-based method was developed. Goal(s): To effectively reconstruct k-space data acquired from the RRFCA, enhancing image quality compared to a conventional stationary array without a scan time penalty. Approach: Conventional GRAPPA was extended by uncovering a subset of the radial grid to cope with the rotation of the RRFCA. Numerical and human brain images were used for validation. Results: Image quality was improved using the proposed method. Up to 58% reduction in RMSE and 2.5% increase in SSIM was achieved while maintaining scan time. Impact: The RRFCA utilising our novel calibration-free, GRAPPA-based,
radial image reconstruction method provides a clinically relevant parallel
imaging technique. In the future, our approach may incorporate compressed
sensing to further reduce motion artifacts, particularly in applications like
cardiac and dynamic MRI. |
09:15 | 0042.
| Fast and motion robust brain examination using simultaneous multi-slice turbo gradient spin echo BLADE Sequence Kun Zhou1, Li Yang1, and Nan Xiao1 1Siemens Shenzhen Magnetic Resonance Ltd., Shenzhen, China Keywords: Motion Correction, Motion Correction Motivation: PROPELLER/BLADE is robust to motion in brain imaging but comes at the cost of longer acquisition time. Goal(s): Our goal was to reduce the acquisition time of commercial sequences (BLADE and EPI) based brain motion-insensitive workflow by a factor of 2. Approach: The SMS-TGSE-BLADE sequence was developed with acceleration techniques, including in-plane GRAPPA, SMS, and EPI readout. Results: Comparable image quality was obtained with the SMS-TGSE-BLADE sequence with more than 2-fold decrease in acquisition time. Impact: The improvement in acquisition speed in the motion-insensitive brain examination (including T1-FLAIR, T2W, T2-FLAIR and DWI) through SMS-TGSE-BLADE sequence may increase patient comfort. It can also increase patient throughput and cost efficiency of healthcare providers. |
09:27 | 0043.
| Deep Learning Reconstruction for Free-Breathing Radial Cine Imaging Mahmut Yurt1, Kanghyun Ryu2, Zhitao Li3, Xucheng Zhu4, Xianglun Mao4, Kawin Setsompop5, Martin Janich4, John Pauly1, Ali Syed5, and Shreyas Vasanawala5,6 1Department of Electrical Engineering, Stanford University, Stanford, CA, United States, 2Korea Institute of Science and Technology, Seoul, Korea, Republic of, 3Department of Radiology, Northwestern University, Chicago, IL, United States, 4GE Healthcare, Stanford, CA, United States, 5Department of Radiology, Stanford University, Stanford, CA, United States, 6Stanford Cardiovascular Institute, Stanford University, Stanford, CA, United States Keywords: Image Reconstruction, Cardiovascular Motivation: We aim to introduce a cardiac cine imaging protocol to address the issues of motion susceptibility and robustness in the previous techniques. Goal(s): Our objective is to demonstrate an accelerated acquisition and high-quality reconstruction framework based on free-breathing radial cardiac cine imaging that shows enhanced patient comfort and robustness against respiratory motion. Approach: We synergistically leverage a raw k-space preprocessing module, region optimized coil compression, and deep learning reconstruction based on memory efficient unrolled neural networks. Results: Our experiments indicate that the proposed framework achieves high reconstruction quality at large acceleration factors (e.g., 8x), in terms of spatial and temporal accuracy. Impact: Conventional cardiac protocols use Cartesian k-space sampling and are susceptible to motion artifacts. We provide an acquisition and reconstruction framework based on a free-breathing protocol and deep learning reconstruction for enhanced patient comfort and robustness against motion artifacts. |
09:39 | 0044.
| Distortion-Free Fat/Water Separated Body Diffusion-Weighted Imaging using Spatio-Temporal Joint Reconstruction Xuetong Zhou1,2, Brian A. Hargreaves1,2,3, and Philip K. Lee1 1Department of Radiology, Stanford University, Stanford, CA, United States, 2Department of Bioengineering, Stanford University, Stanford, CA, United States, 3Department of Electrical Engineering, Stanford University, Stanford, CA, United States Keywords: Pulse Sequence Design, Data Acquisition Motivation: DWI is effective for cancer imaging, but conventional EPI suffers from geometric distortion and chemical shift artifacts. Conventional fat suppression techniques are sensitive to the large B0 and B1+ inhomogeneities in the body. Residual fat causes artifacts and is a confounding factor in using DWI for cancer diagnosis. Goal(s): Perform robust fat/water separation in distortion-free DWI. Approach: A diffusion-weighted EPTI acquisition and joint reconstruction method is used. Separation is performed using chemical shift encoding along the temporal dimension. A distortion-less FSE-based phase navigator is used to resolve shot-to-shot phase. Results: The proposed method is validated in vivo in the brain, head&neck, and breast. Impact: Using the proposed navigated EPTI sequence, we demonstrated fat/water separated DWI that is robust to B0 variation in the body. This will enable more reliable use of DWI to assess cancer and other abnormalities, complementing or replacing contrast-enhanced imaging. |
09:51 | 0045.
| High-fidelity Four-dimensional Abdominal Diffusion-Weighted Imaging Enabled by SCOPER and Multi-Band acceleration (4D-DW-MB-SCOPER) Lu Wang1, Tian Li1, Jing Cai1, and Hing Chiu Chang2 1The Hong Kong Polytechnic University, Hong Kong, China, 2The Chinese University of Hong Kong, Hong Kong, China Keywords: Image Reconstruction, Radiotherapy Motivation: 4D-DWI can benefit the treatment planning in radiotherapy (RT) because of its high tumor-to-tissue contrast. Both 4D-DW-PROPELLER-EPI and 4D-DW-SCOPER have been proposed but suffer from long acquisition time, thereby limiting clinical applications Goal(s): This study aims to develop a new technique to achieve distortion-free 4D-DWI with a practical acquisition time. Approach: 4D-DW-SCOPER and multiband (MB) techniques were combined, termed 4D-DW-MB-SCOPER. In vivo experiments were performed. Results: Results indicate that 4D-DW-MB-SCOPER is feasible for achieving distortion-free 4D-DWI within 7 mins for a coverage of 176 mm in the Superior-Inferior (SI) direction, and has the potential to benefit treatment planning in clinical RT. Impact: The results might offer a new way for clinicians to perform 4D RT planning as well as for patients to have a better treatment outcome. However, the reconstruction time is long for the technique and need to be further investigated. |
10:03 | 0046.
| Image recovery using deep end-to-end posterior networks Jyothi Rikhab Chand1 and Mathews Jacob1 1University of Iowa, Iowa city, IA, United States Keywords: Image Reconstruction, Data Processing Motivation: End-to-End (E2E) trained unrolled algorithms recover MR images with high quality. However, they have large memory demands during training. In addition, these maximum a posteriori methods cannot provide uncertainty estimates. Goal(s): To develop a memory-efficient framework for E2E learning of the posterior probability distribution. Approach: We model the posterior distribution as a combination of the data-consistent-determined likelihood term and the prior, represented using a Convolutional Neural Network whose weights are learned in an E2E fashion using maximum likelihood optimization. Results: The proposed E2E training strategy requires significantly less memory than unrolling. In addition, the model facilitates sampling and provides uncertainty estimates. Impact: The higher memory efficiency of the proposed E2E scheme makes it an attractive option for image reconstruction problems of large dimensions. The learned posterior model provides a minimum mean square estimate and uncertainty maps, which unrolled approaches cannot offer. |