View Presentation Video

Ola Norbeck^1,2, Henric Rydén^1,2, Adam van Niekerk¹, Enrico Avventi^1,2, Tim Sprenger³, Stefan Skare^1,2, Johan Berglund⁴, and Mikael Skorpil ^1,2
1Karolinska Instituet, Stockholm, Sweden, ²Karolinska University Hospital, Stockholm, Sweden, ³MR Applied Science Laboratory Europe, GE Healthcare, Munich, Germany, ⁴Uppsala University Hospital, Uppsala, Sweden

Keywords: Data Acquisition, Data Acquisition, MSK, Dixon, SMS

By combining thin slices, PROPELLER acquisition, SMS acceleration and chemical shift encoding with asymmetrical gradients reformattable pseudo 3D volumes can be acquired. These volumes are, in contrast to 3D RARE volumes, efficiently chemical shift encoded, less susceptible to T2-blurring and less sensitive to motion.

View Presentation Video

Henric Rydén¹, Mikael Skorpil², Matea Borbas³, and Adam van Niekerk^1
1Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden, ²Karolinska Institutet, Stockholm, Sweden, ³Karolinska University Hospital, Stockholm, Sweden

Keywords: Pulse Sequence Design, Fat, Dixon

A novel asymmetric readout waveform for Dixon FSE imaging is presented for efficient sampling, tailored for real-valued estimation of fat/water estimates.

View Presentation Video

Hongyu Li¹, Brendan L. Eck², Mingrui Yang², Jeehun Kim², Ruiying Liu¹, Peizhou Huang³, Dong Liang⁴, Xiaojuan Li², and Leslie Ying^1,3
1Department of Electrical Engineering, University at Buffalo, State University of New York, Buffalo, NY, United States, ²Department of Biomedical Engineering, Program of Advanced Musculoskeletal Imaging (PAMI), Cleveland Clinic, Cleveland, OH, United States, ³Department of Biomedical Engineering, University at Buffalo, State University of New York, Buffalo, NY, United States, ⁴Paul C. Lauterbur Research Center for Biomedical Imaging, Medical AI research center, SIAT, CAS, Shenzhen, China

Keywords: MR Fingerprinting/Synthetic MR, Image Reconstruction

We propose a novel, deep learning-based method, “SuperMRF”, for the reconstruction of MR Fingerprinting (MRF) parametric maps that enables rapid image reconstruction. Built upon a convolutional neural network, SuperMRF uses three loss functions to incorporate additional information from the Bloch equations, estimated maps, de-aliasing, and data consistency losses. We investigate the use of SuperMRF for further acceleration of data acquisition by reducing the number of MRF time frames. Our results demonstrate that proposed SuperMRF is robust to noise and can achieve a 20x reduction in acquired MRF time frames. Tissue property maps can be reconstructed in less than one second.

View Presentation Video

Keerthi Sravan Ravi^1,2, Gautham Nandakumar³, Nikita Thomas³, Mason Lim³, Enlin Qian^2,4, Marina Manso Jimeno^2,4, Pavan Poojar⁵, Zhezhen Jin⁶, Patrick Quarterman⁷, Maggie Fung⁷, Girish Srinivasan³, John Thomas Vaughan Jr.², and Sairam Geethanath^5
1Biomedical Engineering, Columbia University, New York, NY, United States, ²Columbia Magnetic Resonance Research Center, Columbia University, New York, NY, United States, ³PhenoMX, Chicago, IL, United States, ⁴Department of Biomedical Engineering, Columbia University, New York, NY, United States, ⁵Accessible MR Laboratory, Biomedical Engineering and Imaging Institute, Dept. of Diagnostic, Molecular and Interventional Radiology,, Mt. Sinai, New York, NY, United States, ⁶Department of Biostatistics, Columbia University Irving Medical Center, New York, NY, United States, ⁷MR Clinical Solutions, GE, New York, NY, United States

Keywords: Data Acquisition, Data Acquisition

The routine brain screen protocol employed at our institution was accelerated using Look Up Tables to achieve a 1.94x gain in imaging throughput. Image-denoising was performed on accelerated data by leveraging deep learning models trained on contrast-specific publicly available datasets. These were corrupted by native noise during forward modeling. In addition, subject-specific denoising was demonstrated. The superior performance of denoised data on automated volumetry of Alzheimer’s Disease (AD) relevant brain anatomies on T₁w data demonstrated potential for accelerated AD imaging.

Haonan Zhang¹, Qingwei Song¹, Jiazheng Wang², and Ailian Liu^1
1Department of Radiology, the First Affiliated Hospital of Dalian Medical University, Dalian, Dalian, China, ²Philips Healthcare, Beijing, China, Beijing, China

Keywords: Brain Connectivity, Brain

Compared with echo planar imaging diffusion weighted iamging (EPI-DWI), turbo spin echo diffusion weighted imaging (TSE-DWI) can significantly reduce magnetic sensitivity artifacts in skull base imaging. However, the longer scan time limits its clinical promotion. The purpose of this study is to investigate the effect of the compression sensing acceleration factor on the image quality of TSE-DWI in the skull base area.

View Presentation Video

Yang Zhao¹, Yishi Wang², Guangbin Wang¹, and Weibo Chen^2
1Department of Radiology, Shandong Provincial Hospital Affiliated to Shandong First Medical University, Jinan, China, ²Philips Healthcare, Beijing, China

Keywords: Data Acquisition, Data Acquisition, compressed sensing

The magnetization-prepared 2 rapid acquisition gradient echo (MP2RAGE) sequence provides morphological T1-weighted images and quantitative T1 maps. Compressed sensing (CS) acceleration technique has been drawing extensive attention with its high acquisition efficiency for MRI sequence. In this study, volume measurement and ROI analysis were used to assess the images obtained from CS-MP2RAGE sequence under different acceleration factors. We find that CS-MP2RAGE show high similarity in brain structure volumes measurement and ROI analysis with traditional SENSE-MP2RAGE. Those findings prove that CS-MP2RAGE can be considered as an alternative to the SENSE-MP2RAGE, though the specific acceleration factor still needs to be further studied.

shuai hu¹, haonan zhang¹, nan wang¹, qingwei song¹, and ailian liu^1
1the First Affiliated Hospital of Dalian Medical University, dalian, China

Keywords: New Devices, Brain

ACS based cranial MRI has good feasibility in clinical use. Scanning time is reduced by 37%, 40%, and 71% for recommended AF in T1-FLAIR, T2WI, and T2-FLAIR, while the image quality meets the requirement of diagnosis.   

View Presentation Video

Sampada Bhave¹, Saurav Sajib¹, Aniket Pramanik², Mathews Jacob², and Samir Sharma^1
1Canon Medical Research USA Inc, Mayfield, OH, United States, ²University of Iowa, Iowa City, IA, United States

Keywords: Image Reconstruction, Image Reconstruction

Model-based deep learning algorithms offer high quality reconstructions for accelerated acquisitions. Training the regularization parameter λ can lead to instabilities during training. In this work, we evaluated effect of fixing λ parameter while training. We observed no difference in image quality when the network was trained with a fixed λ parameter when the fixed value was equal to the value learned from training. We observed that IQ is dependent on the fixed λ values used during training. Furthermore, we observed that tuning the λ parameter during inference adapts the framework to the SNR of the testing dataset, yielding improved performance. 

View Presentation Video

Brian Bozymski¹, Xin Shen², Ali Caglar Ozen³, Serhat Ilbey³, Albert Thomas⁴, Mark Chiew⁵, William Clarke⁵, Ulrike Dydak^1,6, and Uzay Emir^1,7
1School of Health Sciences, Purdue University, West Lafayette, IN, United States, ²Department of Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, CA, United States, ³Department of Radiology and Radiotherapy, Medical Center - University of Freiburg, Freiburg, Germany, ⁴Department of Radiology, University of California, Los Angeles, CA, United States, ⁵Wellcome Centre for Integrative Neuroimaging, University of Oxford, Oxford, United Kingdom, ⁶Department of Radiology and Imaging Sciences, Indiana University School of Medicine, Indianapolis, IN, United States, ⁷Weldon School of Biomedical Engineering, Purdue University, West Lafayette, IN, United States

Keywords: Data Acquisition, Muscle

Quantitative comparison of novel, rosette trajectory UTE (70 μs) and conventional weighted Cartesian 3D 31P MRSI sequences is performed at 3T in quadriceps muscle.  After previous validation, five healthy volunteers were scanned by both sequences without interruption.  Fitted metabolite maps and SNR calculations of PCr and ATP signals across selected slices and voxels displayed competitive performance between each acquisition.  Retrospective compressed sensing (CS) acceleration results suggest this UTE MRSI may enable faster, higher resolution 31P metabolite mapping in leg muscle and other organs of interest.

View Presentation Video

Maryam Alsameen¹, Zhaoyuan Gong¹, John Laporte¹, Mary Faulkner¹, Mohammad Akhonda¹, and Mustapha Bouhrara^1
1Laboratory of Clinical Investigations, National Institiute on Aging, National Institutes of Health, Baltimore, MD, United States

Keywords: Data Analysis, Brain

We demonstrated the feasibility of compressed sensing (CS) to accelerate myelin water fraction (MWF) imaging using the BMC-mcDESPOT method. Our results showed that derived MWF maps using CS were similar to those derived without CS. Findings from this study indicate that whole brain, high resolution, MWF map can be derived within a few minutes.

View Presentation Video

Yukun Zhang¹, Na Liu¹, Geli Hu², Liangjie Lin², Yanwei Miao¹, and Qingwei Song^1
1the First Affiliated Hospital of Dalian Medical University, Dalian, China, ²Clinical and Technical Support, Philips Healthcare, beijing, China

Keywords: Parallel Imaging, Vessels, compressed sensitivity encoding

The clinical application of phase-contrast magnetic resonance venogram (PC-MRV) is limited by the long scan time. This study aims to accelerate the acquisition of PC-MRV by the compressed sensitivity encoding (CS-SENSE) technique and find the optimal acceleration factor. Results show that the image quality of fast PC-MRV based on CS-SENSE was significantly higher than that based on conventional sensitivity encoding (SENSE) technology, and the CS-SENSE factor 9 was recommended for mild to moderate patients, and CS AF=12 for critical patients.

View Presentation Video

Kang Yan¹ and Craig H Meyer^1,2
1Biomedical Engineering, University of Virginia, Charlottesville, VA, United States, ²Radiology & Medical Imaging, University of Virginia, Charlottesville, VA, United States

Keywords: Image Reconstruction, Relaxometry

A nonconvex low rank regularization (NLR) was proposed to accelerate parameter mapping in the k-p domain. The NLR uses weighted nuclear norm minimization (WNNM) to obtain an optimized solution by differently penalizing singular values, in comparison to traditional low rank methods. The performance of the proposed algorithm was demonstrated for T2 mapping of the kidney. Our study demonstrated that the proposed algorithm outperformed k-p domain-based compressed sensing and L&S algorithms.

View Presentation Video

Bahadir Alp Barlas¹, Kubra Keskin¹, Brian A Hargreaves^2,3,4, and Krishna S Nayak^1,5
1Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, United States, ²Radiology, Stanford University, Stanford, CA, United States, ³Electrical Engineering, Stanford University, Stanford, CA, United States, ⁴Bioengineering, Stanford University, Stanford, CA, United States, ⁵Biomedical Engineering, University of Southern California, Los Angeles, CA, United States

Keywords: Data Acquisition, Low-Field MRI

Metallic implants cause severe distortions in the magnetic field that are best mitigated using multi-spectral acquisitions that require longer scan time. At conventional field strengths, this is compensated by aggressive use of parallel imaging.  However, at low-field strengths (<1.5T), available parallel imaging factors are reduced because body noise dominance prevents the use of dense arrays with smaller elements. Hexagonal sampling (in k_y,k_z space) was previously proposed as a way to achieve 2-fold reduction in imaging time without introducing additional artifacts. In this work, we demonstrate applicability of the hexagonal sampling for 0.55T multi-spectral imaging near metal.

View Presentation Video

Seongwon Na¹, Yousun Ko², Su Jung Ham¹, Mi-Hyun Kim¹, Youngbin Shin¹, Yu Sub Sung¹, Jimi Huh³, Seung Chai Jung¹, and Kyung Won Kim^1
1Asan Medical Center, Seoul, Korea, Republic of, ²University of Ulsan College of Medicine, Seoul, Korea, Republic of, ³Ajou University Hospital, Suwon, Korea, Republic of

Keywords: Machine Learning/Artificial Intelligence, Machine Learning/Artificial Intelligence, Sequence; Classification

For automatic sequence-type classification of brain MRI, we developed the self-supervised machine learning (ML) algorithm, named ImageSort-net, using a rule-based labeling system based on metadata of Digital Imaging and Communications in Medicine (DICOM) image files. Our rule-based labeling system and ImageSort-net showed high classification performance to predict brain MRI sequence type. ImageSort-net showed reliable performance by appending a new dataset to an existing dataset and without human labeling of the whole dataset. This result indicates that sustainable self-learning ML algorithms using the rule-based virtual label in the new datasets are feasible. 

View Presentation Video

Abdul Basit¹, Omair Inam¹, and Hammad Omer^1
1Electrical Engineering, MIPRG, Comsats University Islamabad, Islamabad, Pakistan

Keywords: Parallel Imaging, Cardiovascular, Accelerator, Real-time

GRAPPA is a cartesian pMRI method widely adopted for high-speed image reconstruction in many clinical applications e.g. real-time cardiac MRI. However, general-purpose computers have limited processing capabilities to address the computational complexity of GRAPPA in real-time MRI. Recently, multi-processor system-on-chip (MPSoC) has emerged as a potential candidate to meet the rising computational demands of GRAPPA for real-time image reconstruction. In this paper, design and implementation details of a novel MPSoC based GRAPPA accelerator i.e. SOC-GRAPPA are presented. The experimental results of 18-coil cardiac dataset show that the proposed accelerator is capable of reconstructing 30 frames/second in real-time cardiac MRI. 

View Presentation Video

Xiao Wu¹, Shan Xu², Yao Zhang², Jianzhong Sun², and Peiyu Huang^2
1Radiology, The Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China, ²The Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China

Keywords: Data Acquisition, Machine Learning/Artificial Intelligence, deep-learning ,susceptibility-weighted imaging,magnetic resonance imaging

       In this study, we validated a deep-learning-based method for accelerating susceptibility-weighted imaging (SWI) in 31 clinical subjects. Compared to the fully sampled images, the accelerated SWI images had less noise and imaging artifacts. Although the images had decreased sharpness, the anatomical details of the lesions were mostly kept, and we had not observed false negative\positive lesions. This method could be useful for clinical situations that need timely imaging results.

View Presentation Video

Ajin Joy¹, Uzay Emir^2,3, Paul M. Macey⁴, and M. Albert Thomas^1
1Radiological Sciences, University of California, Los Angeles, Los Angeles, CA, United States, ²School of Health Sciences, Purdue University, West Lafayette, IN, United States, ³Weldon School of Biomedical Engineering, Purdue University, West Lafayette, IN, United States, ⁴School of Nursing and Brain Research Institute, University of California, Los Angeles, Los Angeles, CA, United States

Keywords: Data Acquisition, Brain

Undersampling spatial and  spectral dimensions is essential to achieve clinically feasible scan times in multi-dimensional spectroscopic imaging. Sampling pattern of rosette trajectory has higher incoherence than that of the other non-Cartesian trajectories like spiral and radial, and can achieve higher compressed-sensing reconstruction performance. While rosette spectroscopic imaging has been attempted for 2D (2 spatial+1 spectral) and 3D (3 spatial+1 spectral) spectroscopic imaging, it has thus far not been shown for J-resolved spectroscopic imaging. In this pilot study, we implemented a rosette echo-planar J-resolved spectroscopic imaging (ROSE-JRESI) sequence and studied the feasibility of high acceleration factors for clinically feasible runtimes.

View Presentation Video

Rui Tian¹, Theodor Steffen¹, and Klaus Scheffler^1,2
1High-Field MR center, Max Planck Institute for Biological Cybernetics, Tuebingen, Germany, ²Department for Biomedical Magnetic Resonance, University of Tübingen, Tuebingen, Germany

Keywords: New Trajectories & Spatial Encoding Methods, New Trajectories & Spatial Encoding Methods, nonlinear gradient encoding

We further developed a recent idea called spread spectrum MRI to reduce the sampling time by rapidly modulating spins with localized magnetic fields during signal readout. Given phantom experiments tested and safety evaluation for human subjects performed, this time, we started in-vivo measurements of human head with multi-slice FLASH sequence accelerated by local B₀ coil modulations, and examined the reconstructed image quality. Sinusoidal modulation schemes in various phase offset patterns and frequencies given different scanner bandwidth were tested and compared, which were shown to boost the image acceleration from 6-fold (i.e., SENSE only) to about 8-fold in one phase-encoded dimension.

View Presentation Video

Merry Mani¹, Xinzeng Wang², and Baolian Yang^3
1University of Iowa, Iowa City, IA, United States, ²GE Healthcare, Houston, TX, United States, ³GE Healthcare, Waukesha, WI, United States

Keywords: Brain Connectivity, High-Field MRI

Achieving high resolution in diffusion MRI is challenging due to its inherently low SNR, vulnerability to motion and other EPI-related artifacts. On higher field-strengths, the SNR advantage can be exploited to push the resolution if the echo-time can be effectively reduced and the increased number of slices can be efficiently acquired. Combining acceleration techniques such as multi-band and parallel imaging is critical for this approach. Here we combine two deep-learned reconstruction priors, one pertaining to the q-space and another pertaining to image artifact removal, into a model-based iterative reconstruction framework to improve the quality of highly accelerated high-resolution 7T DWIs.

View Presentation Video

Jonghyo Youn¹, Juhyung Park¹, Sooyeon Ji¹, Hongjun An¹, Hwan Heo², MyeongOh Lee², Soohwa Song², Eung Yeop Kim³, and Jongho Lee^1
1Seoul National University, Seoul, Korea, Republic of, ²Heuron Co.Ltd., Incheon, Korea, Republic of, ³Radiology, Samsung Medical Center, Seoul, Korea, Republic of

Keywords: Image Reconstruction, Parkinson's Disease

 Nigral hyperintensity detection in substantia nigra is a potential biomarker for PD. An advanced SWI method, SMWI, has demonstrated reliable detection of the hyperintensity at 3T but it requires a 4 m 15 s scan protocol which is too long for PD patients, suffering from motion artifacts. In this study, we developed a new 2 m 56 s protocol by reducing phase FOV and applying deep learning-powered denoising to compensate for SNR loss from the shorter scan. The new protocol was validated using both simulated and real data via an automated tool and an expert radiologist.