| Computer # |
| 3562.
 | 17 | Unlocking Data-Consistent Synthesis of Clinical Contrasts from Magnetic Resonance Fingerprinting with Semi-Supervised Learning Mahmut Yurt1, Zihan Zhou2, Congyu Liao3, Cagan Alkan1, Xiaozhi Cao3, Nan Wang3, Julio Oscanoa4, Sophie Schauman5, Mengze Gao6, Zhitao Li7, Tolga Cukur8, Bruno Soares3, Ali Syed3, Shreyas Vasanawala3, John Pauly1, and Kawin Setsompop3
1Department of Electrical Engineering, Stanford University, Stanford, CA, United States, 2Stanford University, Stanford, CA, United States, 3Department of Radiology, Stanford University, Stanford, CA, United States, 4Department of Bioengineering, Stanford University, Stanford, CA, United States, 5Karolinska Institutet, Solna, Sweden, 6Department of Biomedical Physics, Stanford University, Stanford, CA, United States, 7Department of Radiology, Northwestern University, Chicago, IL, United States, 8Department of Electrical and Electronics Engineering, Bilkent University, Ankara, Turkey Keywords: MR Fingerprinting, MR Fingerprinting Motivation: We aim to introduce an end-to-end, ultra-fast acquisition and synthesis protocol for contrast-weighted image generation from magnetic resonance fingerprinting. Goal(s): Our objective is to facilitate data compilation from large patient cohorts to enhance generalizability and accuracy of synthesis. Approach: We leverage a semi-supervised framework to enable model training using highly-accelerated ground-truth data of the target contrasts, and introduce data-consistent synthesis in inference by performing subject-specific fine-tuning and validation. Results: Our experiments indicate that the proposed method enables high-quality synthesis using network models trained on prospectively undersampled data of the contrast-weighted images. We show that data-consistent synthesis helps improve synthesis quality and mitigate hallucinations. Impact: Conventional synthesis models for MRF perform a single-shot inference and are prone to hallucinations and inaccuracy. We introduce semi-supervised learning that enables data-consistency in inference, by merely getting an additional, ultra-fast 1-2min data of the target contrasts for subject-specific fine-tuning. |
| 3563.
 | 18 | TransUnet Based Deep Learning with Tissue-weighted Loss Function for Accelerated Magnetic Resonance Fingerprint Reconstruction Jintao Wei1,2, Zihan Zhou3,4, Bo Dong1,2, Paween Wongkornchaovalit1,2, HuiHui Ye5, and Hongjian He1,6,7
1Center for Brain Imaging Science and Technology, Zhejiang University, Hangzhou, China, 2College of Biomedical Engineering and Instrument Science, Zhejiang University, Hangzhou, China, 3Department of Radiology, Stanford university, Stanford, CA, United States, 4Graduate School of Education, Stanford university, Stanford, CA, United States, 5State Key Laboratory of Modern Optical Instrumentation, College of Optical Science and Engineering, Zhejiang University, Hangzhou, China, 6School of Physics, Zhejiang University, Hangzhou, China, 7State Key Laboratory of Brain-Machine Intelligence, Zhejiang University, Hangzhou, China Keywords: MR Fingerprinting, Image Reconstruction, transformer, MRF, reconstruction, reproducibility Motivation: MRF data with short time frames make dictionary matching challenging with large quantitative errors. Goal(s): We proposed a deep learning method by taking account in global Information and tissue-specific relaxometry, for better quantitative results using shorter frames than dictionary matching with good reproducibility. Approach: We proposed to combine Unet with an embedded transformer and a tissue-weighted loss function. We compared our approach to dictionary matching and examined various levels of reduced scan time. Results: The proposed method has less error with shorter frame data comparing dictionary matching. This is more effective for the gray matter and for scan times shorter than 300 frames. Impact: The proposed method
improves the problem of poor quality of reconstructed images for dictionary
matching in shorter frames, in addition this method has good repeatability and
provides a powerful reconstruction algorithm for MRF sequence acceleration. |
| 3564.
 | 19 | Deep Structure-Preserved Graph Embedding for Improved MRF Reconstruction Peng Li1, Yuping Ji1, and Yue Hu1
1Harbin Institute of Technology, Harbin, China Keywords: MR Fingerprinting, Image Reconstruction, Graph Embedding, Manifold Learning, structure-preserved, Deep Unrolling Motivation: Improve MRF Reconstruction. Goal(s): Introduce a novel deep-learning framework based on the structure-preserved graph embedding for improved MRF reconstruction. Approach: We propose a reconstruction framework based on graph embedding, modeling the high-dimensional MRF data and the parameter maps as graph data nodes. To improve the accuracy of the estimated graph structure and the computational efficiency of the proposed framework, we unroll the iterative steps into a deep neural network and introduce a learned graph embedding module to adaptively learn the graph structure. Results: Numerical experiments demonstrate that our approach can reconstruct high-quality parameter maps within reduced computational cost. Impact: By redefining the MRF reconstruction problem as a structure-preserved graph embedding problem, the proposed method can effectively reduce the computational complexity of MRF reconstruction compared to data-priors-driven methods. |
| 3565.
 | 20 | Cluster-Based Low-Rank Regularization for Cardiac Rosette MR Fingerprinting Evan Cummings1,2, Gastao Cruz1, Sydney Kaplan1,2, Jacob Richardson1, Jesse Hamilton1,2, and Nicole Seiberlich1,2
1Radiology, University of Michigan, Ann Arbor, MI, United States, 2Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States Keywords: MR Fingerprinting, MR Fingerprinting, Sparse & Low-Rank Models, Heart, Quantitative Imaging Motivation: Breathholds limit the amount of data which can be acquired in cardiac MRF, which can impact the precision of fat/water separated T1, T2, and T2* maps. Goal(s): We developed a regularization method to reconstruct accurate maps from multi-echo cMRF data without introducing blurring into the resulting tissue property maps. Approach: A k-means cluster-based approach is used to group the signal evolutions during reconstruction and a low-rank constraint is applied to each cluster. We compared our method to existing approaches in 23 healthy volunteers. Results: This approach can be used to generate accurate myocardial T1, T2, and T2* maps using rosette MRF data. Impact: Traditional cardiac MRF reconstructions can fail when
working with multi-echo rosette MRF data due to insufficient sampling. We developed
a reconstruction method which enables T1, T2, and T2*
maps to be collected in a single breathhold without compromising accuracy. |
| 3566.
 | 21 | Factorized Spatiotemporal Convolutions for Simultaneous Multislice Magnetic Resonance Fingerprinting Lan Lu1, Yilin Liu2, Amy Zhou1, Pew-Thian Yap3, and Yong Chen4,5
1Radiation Oncology, Cleveland Clinic, Cleveland, OH, United States, 2Computer Science, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, 3Radiology, University of North Carolina at Chapel Hill, Chapel Hill, NC, United States, 4Radiology, Case Western Reserve University, Cleveland, OH, United States, 5Radiology, University Hospitals Cleveland Medical Center, Cleveland, OH, United States Keywords: MR Fingerprinting, Quantitative Imaging Motivation: Quantitative MR imaging with volumetric coverage is relatively slow, which hinders its clinical evaluation. Goal(s): To leverage simultaneous multislice MR Fingerprinting (SMSMRF) and deep learning to achieve high multi-band factors for rapid quantitative imaging. Approach: We introduced a Spatio-Temporal UNet (STUN) method to exploit both spatial and temporal correlations of signals in SMSMRF to largely accelerate data acquisition. Results: High multi-band factors (3 and 4) with an additional 4x acceleration along the temporal dimension were achieved, providing rapid data sampling of 1.5~2 sec per slice for quantitative brain imaging. Impact: The developed SMSMRF method holds great
potential to accelerate quantitative imaging for challenging subjects, f.g.
pediatric patients. |
| 3567.
 | 22 | Improved 3D Prostate MR Fingerprinting with Cross-Domain Spatio-Temporal Reconstruction Network Jae-Yoon Kim1, Jae-Hun Lee1, Dongyeob Han2, Moon Hyung Choi3, and Dong-Hyun Kim1
1Department of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea, Republic of, 2Siemens Healthineers Ltd, Siemens Korea, Seoul, Korea, Republic of, 3Department of Radiology, Eunpyeong St. Mary’s Hospital, College of Medicine, The Catholic University of Korea, Seoul, Korea, Republic of Keywords: MR Fingerprinting, MR Fingerprinting Motivation: 3D Magnetic Resonance Fingerprinting is time-consuming, requiring full-time measurements. Shortening scan time while maintaining data quality enhances MRF's clinical utility. Goal(s): Our goal was to develop reconstruction process for prostate MRF based on the neural network. This approach aims to improve image quality and parameter map accuracy. Approach: We introduced the neural network composed of a combination of CNN and ANN and utilized compressed dictionary, enabling efficient cross-domain utilization of information. Results: Our approach enhances quality and accuracy of generated parameter maps, demonstrating the potential to expedite MRF scans for prostate imaging. Impact: Our proposed scheme for accelerated MRF
reconstruction can improve quantitative imaging, thus providing faster and more
accurate prostate diagnosis and treatment. This development has positive impacts
on patient care, reduces scanning times, and promotes additional research in
medical image reconstruction. |
| 3568.
 | 23 | Rapid Multiparametric Quantitative Bilateral Breast MR Fingerprinting Using a Phase-Sensitive Deep Learning Network. Jessica A. Martinez1, Elizabeth J. Sutton2, Ricardo Otazo1, and Ouri Cohen1
1Medical Physics, Memorial Sloan Kettering Cancer Center, New York,, NY, United States, 2Radiology, Memorial Sloan Kettering Cancer Center, New York,, NY, United States Keywords: MR Fingerprinting, MR Fingerprinting, Quantitative Susceptibility Mapping Motivation: To rapidly obtain MRF-based multiparametric quantitative maps in the breast. Goal(s): To explore the feasibility of using PS-DRONE and an EPI-MRF sequence in breast imaging to simultaneously quantify T1, T2, PD, B1, phase, and QSM maps. Approach: MRF data were acquired at 3T. Tissue parameters were reconstructed using PS-DRONE, including QSM maps computed from the estimated phase. Results: Bilateral breast T1, T2, PD, B1 and phase (for QSM analysis) maps were obtained using PS-DRONE and an EPI-MRF sequence. Scan time was 2 minutes and 30 seconds. Parameter reconstruction time was one second. Maps presented differences between the two breasts consistent with diffusion images. Impact: Comprehensive quantitative T1, T2, PD, B1 and phase
(QSM) bilateral breast imaging in under 2.5 minutes can potentially improve the detection and characterization
of breast cancer and treatment response in a clinical setting without the use
of a contrast agent. |
| 3569.
 | 24 | MR Vascular Fingerprinting with Deep Learning to Estimate Brain Physiological Parameters Chieh-Te Lin1, Gregory J. Wheeler1, and Audrey P. Fan1,2
1Biomedical Engineering, University of California, Davis, DAVIS, CA, United States, 2Neurology, University of California, Davis, DAVIS, CA, United States Keywords: MR Fingerprinting, MR Fingerprinting Motivation: Magnetic resonance vascular fingerprinting quantitatively measures microvascular blood oxygen saturation, cerebral blood volume, and vessel radii. Matching simulated signals to in-vivo data is computationally expensive, therefore, we leverage deep learning to alleviate the burden. Goal(s): Build a model to simultaneously and accurately estimate three physiological parameters from a GESFIDE (gradient echo sampling of free induction decay and echo) sequence. Approach: The model has two fully-connected layers to estimate three parameters, and was validated with synthetic signals and healthy subject parameter mapping. Results: The model achieves comparable root-mean-squared-error to traditional fingerprint matching in test signals. We show physiological reasonable values in healthy subject maps. Impact: We leverage deep learning in MR vascular fingerprinting to simultaneously
estimate brain physiological parameters through training with simulated
vascular dictionaries. The model enables quantitative measurements of oxygenation,
blood volume and vessel radius in test signals and in-vivo mapping. |
| 3570.
 | 25 | DAES: Self-Supervised Parameter Estimation Model for MR Fingerprinting Jinghang Tan1, Huihui Ye2, Mengze Gao3, Zihan Li4, Qiyuan Tian4, and Berkin Bilgic5,6
1School of Software, Tsinghua University, Beijing, China, 2State Key Laboratory of Modern Optical Instrumentation, College of Optical Science and Engineering, Zhejiang University, Hangzhou, China, 3Stanford University, Stanford, CA, United States, 4Department of Biomedical Engineering, Tsinghua University, Beijing, China, 5Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, 6Harvard Medical School, Boston, MA, United States Keywords: MR Fingerprinting, MR Fingerprinting, self-supervised learning Motivation: Accurate estimation of relaxation parameters using MRF requires lengthy acquisitions as it benefits from having multiple spiral interleaves to boost the data quality. Goal(s): We aim to reduce acquisition time by denoising highly under-sampled data while retaining the fidelity of the estimated parameter maps. Approach: An unsupervised convolutional neural network called DAES is proposed. It combines Denoising Auto-coder (DAE) with subspace modeling, taking advantage of both denoising framework and Bloch simulation-based dictionary information. Results: DAES outperforms conventional dictionary matching in both simulated and in-vivo data for MRF, demonstrating stronger ability to estimate parameters from highly under-sampled data. Impact: Magnetic Resonance Fingerprinting with the proposed unsupervised Denoising Auto-encoder permits high-quality T1 and T2 mapping while substantially reducing the acquisition time. |
| 3571.
 | 26 | Towards Vector Encoding Methods on GPUs for Dictionary Search Memory Optimization Gabriel Zihlmann1, Najat Salameh1, and Mathieu Sarracanie1
1Center for Adaptable MRI Technology (AMT Center), Institute of Medical Sciences, School of Medicine, Medical Sciences & Nutrition, University of Aberdeen, Aberdeen, United Kingdom Keywords: MR Fingerprinting, Quantitative Imaging, Dictionary Search, GPU Motivation: GPU memory size can be a limiting factor for MRF dictionary search. Accelerated grouped search on GPUs was implemented previously without addressing GPU memory constraints.
Goal(s): To investigate the feasibility of GPU memory footprint reduction through vector storage format compression.
Approach: We built a CPU-only prototype search engine storing dictionaries as approximations enabling 2x and 4x memory reduction. Additionally, we implemented a post processing step that refines approximate results using the non-approximated vectors.
Results: Refinement was effective at mitigating errors from vector compression. Overall, 2x group memory compression resulted in no quality loss and speed gains, while 4x compression resulted in speed loss.
Impact: Dictionary
search errors from vector storage compression of up to 4x have been found to be
well mitigated by an uncompressed refinement step. This indicates the
feasibility of implementing compression on GPUs to make efficient use of
limited GPU memory. |
| 3572.
 | 27 | Cardiac MR Fingerprinting with a Low-Rank Reconstruction for Simultaneous T1, T2, and T1ρ Mapping Sydney Kaplan1,2, Gastao Lima da Cruz2, Jesse Hamilton1,2, and Nicole Seiberlich1,2
1Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States, 2Radiology, University of Michigan, Ann Arbor, MI, United States Keywords: MR Fingerprinting, MR Fingerprinting, Cardiac MRI, Quantitative MRI, Sequence Design, Contrast Mechanisms Motivation: Quantification of T1, T2, and T1ρ may provide insight into myocardial fibrosis without the need for gadolinium-based contrast agents. Goal(s): This study seeks to present and validate a rapid, single breath-hold cMRF approach for simultaneously mapping T1, T2, and T1ρ. Approach: A cardiac-gated MRF sequence was designed to measure T1, T2, and T1ρ and was tested in simulation, phantom, and six healthy subjects. Results: The proposed T1/T2/T1ρ cMRF technique yielded accurate high-resolution T1, T2, and T1ρ maps in simulation, phantom, and in vivo. Impact: Cardiac Magnetic Resonance Fingerprinting
can be used to quickly, accurately, and simultaneously map T1, T2,
and T1ρ, and has the potential to quickly probe myocardial fibrosis
without contrast. |
| 3573.
 | 28 | Creation of Voxelwise 2D Lookup Tables (LUTs) for MRF-based Synthesis of Qualitative Images Andrew Dupuis1, Yong Chen2, Mark A Griswold1,2, and Rasim Boyacioglu2
1Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States, 2Radiology, School of Medicine, Case Western Reserve University, Cleveland, OH, United States Keywords: MR Fingerprinting, Visualization Motivation: Address the challenge of integrating MRF data into existing MRI analysis via synthetic images without introducing spatial artifacts or hallucinations possible with CNNs. Goal(s): Generate static lookup tables (LUTs) mapping from T1/T2 value space directly to grayscale visualizations matching clinical contrasts. Approach: A simple pixel-wise regression network was trained on a public dataset of MRF data and weighted images. Static LUTs were generated from dictionaries of T1/T2 combinations, then applied to MRF-derived maps for visualization and processing via FSL. Results: Successful generation of synthetic contrast LUTs ensures reproducibility and allows instantaneous visualization or registration of MRF maps in a more conventional grayscale format. Impact: Integration of MRF into traditional analysis pipelines suffers because quantitative maps have inherently different contrasts from weighted images. LUTs for instant, deterministic generation of weighted contrasts from T1/T2 maps allow for direct use of tools like FSL with MRF data. |
| 3574.
 | 29 | Trade-off between Readout Duration and Concomitant Field Effect in 3D MRF at 0.55T Zhibo Zhu1, Nam G. Lee2, and Krishna S. Nayak1
1Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, United States, 2Afred E. Mann Department of Biomedical Engineering, University of Southern California, Los Angeles, CA, United States Keywords: MR Fingerprinting, MR Fingerprinting Motivation: The use of long readouts in 3D MRF improves SNR efficiency but also suffers from stronger concomitant field effects. Goal(s): To evaluate the trade-off between readout duration and concomitant field effects in 0.55T 3D MRF and evaluate mitigation using the MaxGIRF framework. Approach: We performed 0.55T 3D MRF scans of 14 readout durations on the NIST/ISMRM phantom and compared reconstructions with and without concomitant field effects compensation. We analyzed T1 array's standard deviations as a function of readout duration. Results: With correction, MRF T1 standard deviation reduced to stability due to improved SNR efficiency of long readouts, however, broke due to uncorrectable blurring. Impact: Practical trade-off between a readout duration and concomitant field effects are demonstrated. 0.55T 3D MRF benefits from a longer readout duration (6.1-14.7ms) provided a concomitant field effect correction. |
| 3575.
 | 30 | Temporal Low-Rank based k-space Sampling Pattern Optimization for MR Fingerprinting Felix Horger1,2, Sarah McElroy1,2,3, Joseph Hajnal1,2,4, and Shaihan Malik1,2,4
1School of Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, 2London Collaborative Ultra high field System, London, United Kingdom, 3MR Research Collaborations, Siemens Healthcare Limited, Camberley, United Kingdom, 4Centre for the Developing Brain, London, United Kingdom Keywords: MR Fingerprinting, MR Fingerprinting Motivation: MRF can estimate tissue parameters with high efficiency, requiring optimization of sequence parameters alongside k-space sampling patterns. A comprehensive optimization framework was not established yet. Goal(s): Develop a framework for optimizing k-space sampling and understanding reconstruction errors for MRF using temporal low-rank reconstruction. Approach: We quantify MRF performance with the condition number of temporal low-rank system matrices and show suitability in simulation and phantom experiments. Results: We derive optimality-criteria for schedule and sampling, and provide an algorithm for sampling optimization. We demonstrate that systematic deviations from the signal model are a major source of errors in MRF, and address these with center-weighted sampling. Impact: Our results are relevant for researchers interested in the fundamental understanding of MR Fingerprinting. Our theory helps designing MRF sequences, guiding future aspirations to jointly optimize sampling and flip-angle schedule, and identifying significant sources of errors in existing implementations. |
| 3576.
 | 31 | Mitigating MR fingerprinting undersampling errors is more effective through sequence optimization than via low rank reconstruction Martijn Nagtegaal1,2, Frans Vos2,3, Matthias J.P. van Osch1, and David G.J. Heesterbeek4,5
1C.J. Gorter MRI Center, Radiology, Leiden University Medical Center, Leiden, Netherlands, 2Department of Imaging Physics, Delft University of Technology, Delft, Netherlands, 3Department of Radiology and Nuclear Medicine, Erasmus MC, Rotterdam, Netherlands, 4Computational Imaging Group for MR Therapy and Diagnostics, University Medical Center Utrecht, Utrecht, Netherlands, 5Department of Radiotherapy, University Medical Center Utrecht, Utrecht, Netherlands Keywords: MR Fingerprinting, MR Fingerprinting Motivation: To reduce undersampling errors in MR Fingerprinting maps, both sequence optimization and low rank (LR) reconstruction methods have been proposed , but these have not been compared or combined. Goal(s): To compare the effectiveness of LR-reconstruction and sequence optimization for reducing undersampling errors. Approach: Undersampled 2D spiral MRF data as acquired with 7 different flip angle patterns in 6 healthy subjects are reconstructed with the standard vendor reconstruction and an in-house LR-reconstruction and compared to fully-sampled MRF T1- and T2-maps. Results: Undersampling-optimized sequences showed a reduced error compared to other sequences, even after LR-reconstruction. Impact: We show
that optimized flip-angle patterns with or without LR-reconstruction outperform
traditional schemes with LR reconstruction. Inclusion of LR-reconstruction in
the optimization step is not essential, when designing sequences that minimize
undersampling artifacts. |
| 3577.
 | 32 | Robust Highly-accelerated MR Fingerprinting Using Transformer-based Deep Learning Peizhou Huang1, Brendan Eck2, Ruiying Liu3, Hongyu Li3, Mingrui Yang2, Jeehun Kim2, Xiaoliang Zhang1, Xiaojuan Li2, and Leslie Ying1,3
1Biomedical Engineering, State University of New York at Buffalo, Buffalo, NY, United States, 2Lerner Research Institute, Cleveland Clinic, Cleveland, OH, United States, 3Electrical Engineering, State University of New York at Buffalo, Buffalo, NY, United States Keywords: MR Fingerprinting, MR Fingerprinting Motivation: MR fingerprinting (MRF) conventional reconstruction methods need a substantial reconstruction time and memory space. We aim to propose a novel deep-learning method for accelerated MRF reconstruction. Goal(s): To achieve more accurate quantification reconstruction for T1 and T2 from highly undersampled MRF data. Approach: A novel training process was also proposed to construct reliable training data with noise-like aliasing artifacts boosted by Transformer network without need to know the structure information. Results: Experimental results demonstrate that the proposed method achieves more accurate quantification for T1 and T2 than pattern matching and DRONE. Impact: The proposed method can
generate more accurate quantitative maps for highly accelerated MRF data that
enable clinical use in real application. In addition, the proposed training
process is robust to different structures in the image to be reconstructed. |