Hongli Fan^1,2, Lisa Bunker³, Alexandra Zezinka Durfee³, Xiaohong Joe Zhou⁴, Argye E. Hillis³, and Hanzhang Lu^1,2,5
1Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, United States, ²The Russell H. Morgan Department of Radiology & Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, United States, ³Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, MD, United States, ⁴Center for Magnetic Resonance Research and Department of Radiology, University of Illinois at Chicago, Chicago, IL, United States, ⁵F. M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, MD, United States

Quantitative mapping of brain perfusion, diffusion, T₂^*, and T₁ has a broad range of clinical applications. The present work aims to develop a novel pulse sequence that can simultaneously map perfusion, diffusion, T₂^* and T₁ with MR fingerprinting, dubbed MRF-PDT, with a total scan time of <6 minutes. This technique was first demonstrated on healthy volunteers, then on two patients with ischemic stroke. All maps derived from MRF-PDT exhibited the expected image contrasts with quantitative values consistent with those reported in literature. In addition, test-retest studies confirmed the reproducibility of the proposed technique.

Walter Zhao^1,2, Zheyuan Hu¹, Anahita Fathi Kazerooni^3,4, Gregor Körzdörfer⁵, Matthias Nittka⁵, Christos Davatzikos^3,4, Satish E. Viswanath¹, Xiaofeng Wang⁶, Chaitra Badve⁷, and Dan Ma^1
1Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States, ²Medical Scientist Training Program, Case Western Reserve University, Cleveland, OH, United States, ³Center for Biomedical Image Computing and Analysis, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, United States, ⁴Department of Radiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, United States, ⁵Siemens Healthineers, Erlangen, Germany, ⁶Department of Quantitative Health Sciences, Cleveland Clinic, Cleveland, OH, United States, ⁷Department of Radiology, Case Western Reserve University and University Hospitals Cleveland Medical Center, Cleveland, OH, United States

MR fingerprinting (MRF) is a rapid, quantitative imaging approach with significant potential for use in clinical studies, including radiomic applications. Due to its quantitative nature, robustness to system imperfections, and requiring fewer image preprocessing steps, we believe MRF radiomics is uniquely positioned to offer improved reproducibility and generalizability compared to conventional MRI. Here we report reproducibility results of MRF T1 and T2 radiomic features in the healthy human brain, and introduce a novel physics-informed quantization approach for improved reproducibility of quantitative image texture features.

Shohei Fujita^1,2, Yujiro Otsuka^1,3,4, Katsutoshi Murata⁵, Gregor Koerzdoerfer⁶, Mathias Nittka⁶, Yumiko Motoi^7,8, Madoka Nakajima^8,9, Koji Murakami¹⁰, Issei Fukunaga¹, Koji Kamagata¹, Osamu Abe², and Shigeki Aoki^1
1Department of Radiology, Juntendo University, Tokyo, Japan, ²Department of Radiology, The University of Tokyo, Tokyo, Japan, ³Milliman Inc, Tokyo, Japan, ⁴Plusman LLC, Tokyo, Japan, ⁵Siemens Healthcare Japan KK, Tokyo, Japan, ⁶Siemens Healthcare GmbH, Erlangen, Germany, ⁷Department of Neurology, Juntendo University, Tokyo, Japan, ⁸Medical Center for Dementia, Juntendo University, Tokyo, Japan, ⁹Department of Neurosurgery, Juntendo University, Tokyo, Japan, ¹⁰Division of Nuclear Medicine, Department of Radiology, Juntendo University, Tokyo, Japan

We developed a framework utilizing MR fingerprinting and a complex-valued neural network to detect brain amyloid burden. The tailored neural network was trained on real amyloid-PET imaging data and MR fingerprinting acquisitions to estimate PET-derived amyloid deposition from the MR fingerprinting signal evolutions. This complex-valued neural network architecture, designed to increase sensitivity to amyloid-related signals, showed a subject-level amyloid positivity classification with AUC = 0.87 in patients with cognitive decline. The proposed method enables non-invasive amyloid burden mapping, T1 and T2 mapping in a single scan, and is suitable not only for diagnosis but also for monitoring amyloid-reducing treatments.

Quentin Dessain¹, Nicolas Delinte¹, Benoit Macq¹, and Gaëtan Rensonnet^1
1ICTEAM, Université Catholique de Louvain, Louvain La Neuve, Belgium

We estimate microstructural features of crossing fascicles in the white matter by using a fast multi-compartment fingerprinting, an extension of MR fingerprinting to diffusion MRI. The acceleration uses efficient sparse optimization and a dedicated feed-forward neural network to circumvent the inherent combinatorial complexity of the fingerprinting estimation. The accuracy of the results and the speedup factors obtained on in vivo brain data suggest the potential of our method for a fast quantitative estimation of microstructural features in complex white matter configurations.

David Heesterbeek¹, Martin van Gijzen², Frans Vos^1,3, and Martijn Nagtegaal^1
1Department of Imaging Physics, Delft University of Technology, Delft, Netherlands, ²Department of Numerical Analysis, Delft University of Technology, Delft, Netherlands, ³Department of Radiology and Nuclear Medicine, Erasmus MC, Rotterdam, Netherlands

In MR Fingerprinting acquisitions, the choice of flip angle sequence has a strong effect on the accuracy of the estimated parameter maps. Undersampling artefacts in time-series images are a dominant source of error for standard MRF estimations. We propose to use an undersampling-error model leveraging on perturbation theory, to optimise flip angle patterns taking into account the k-space readout, a realistic ground truth and signal phase. In vivo scans show strong visible reduction in error. Root-mean-square errors in $$$T_1$$$ and $$$T_2$$$ were reduced with at least 6.4%-points and 9%-points respectively, compared to a Cramér-Rao bound optimised and a conventional pattern.

Christian Guenthner¹, Johanna Stimm¹, and Sebastian Kozerke^1
1ETH and University Zurich, Zurich, Switzerland

In Magnetic Resonance Fingerprinting, the accuracy of the results is dominated by undersampling artifacts. While in classical relaxometry techniques, the omission of data always leads to a larger error, in fingerprinting, undersampling artifacts can lead to both an increase or a decrease in error. The “temporal encoding efficiency” of fingerprinting can be analyzed based on the change in matching error upon omission of a single time point (leave-one-out). We propose a first-order perturbation of the undersampling error to visualize and identify temporal sequence segments of primary parameter encoding and apply these insights to shorten an exemplary MRF sequence by truncation.

Ouri Cohen¹, Robert Young², Christian T Farrar³, and Ricardo Otazo^1
1Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, United States, ²Radiology, Memorial Sloan Kettering Cancer Center, New York, NY, United States, ³Radiology, Athinoula A. Martinos Center, Charlestown, MA, United States

CEST imaging is a promising tool for diagnosis and evaluation of treatment response in tumors. However, conventional CEST is not quantitative and requires long acquisition times. A recently developed technique, CEST MR fingerprinting (CEST-MRF), overcomes many of the technical limitations of conventional CEST but still suffers from limited volumetric coverage. In this work, we propose a novel multi-slice CEST-MRF pulse sequence and deep learning reconstruction method to enable volumetric coverage without the need for additional scan time. Numerical simulations and in vivo experiments in a healthy subject are performed to demonstrate feasibility and utility of the proposed multi-slice CEST-MRF technique.

Evan Cummings^1,2, Yuchi Liu², Yun Jiang^1,2, Kathleen Ropella-Panagis², Jesse Hamilton^1,2, and Nicole Seiberlich^1,2
1Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States, ²Radiology, University of Michigan, Ann Arbor, MI, United States

In this work, Rosette MRF was modified and deployed to simultaneously map T₁, T₂, and T₂*. Mapping is demonstrated on the ISMRM/NIST phantom for verification of quantitative accuracy, and on four healthy volunteers to demonstrate potential feasibility in vivo.

Daniel West^1,2, Raphael Tomi-Tricot³, Lucilio Cordero-Grande^1,4,5, Pip Bridgen¹, Tom Wilkinson¹, Sharon Giles¹, Joseph Hajnal^1,4, and Shaihan Malik^1,4
1Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, ²London Collaborative Ultra High Field System (LoCUS), London, United Kingdom, ³MR Research Collaborations, Siemens Healthcare Limited, Frimley, United Kingdom, ⁴Centre for the Developing Brain, King's College London, London, United Kingdom, ⁵Biomedical Image Technologies, ETSI Telecomunicación, Universidad Politécnica de Madrid & CIBER-BNN, Madrid, Spain

Since semisolid tissue components are strong determinants of relaxation parameters, estimation of semisolid properties may lead to more robust and direct characterization of tissues than relaxation measurement. We have used multiband MR fingerprinting at ultra-high field in an exploratory study towards high resolution quantitative imaging with efficient parameter encoding. To achieve this, two variants of a steady-state sequence are proposed that employ different cycling of nonselective multiband RF pulses: one to highlight dipolar order effects apparent in lipid bilayer-like materials such as myelin, and the other to measure semisolid T₂. Initial phantom results and preliminary in vivo investigation are presented.

Yong Chen¹, Rasim Boyacioglu¹, Gamage Sugandima Nishadi Weragoda², Michael Martens², Mark Griswold¹, and Chaitra Badve^1,3
1Radiology, Case Western Reserve University, Cleveland, OH, United States, ²Physics, Case Western Reserve University, Cleveland, OH, United States, ³Radiology, University Hospitals Cleveland Medical Center, Cleveland, OH, United States

In this pilot study, we aim to analyze MR Fingerprinting (MRF) signal using deep learning network to assess the performance of tissue classification in gliomas. A U-Net based convolutional neural network was trained to learn glioma grades based on the SVD-compressed fingerprint acquired using MRF. Based on data acquired from a 5-minute MRF scan, the method shows great potential to accurately classify glioma grades without the need of image registration and contrast administration.

Matteo Cencini^1,2, Marta Lancione^1,2, Laura Biagi^1,2, Rossella Pasquariello¹, Luca Peretti^2,3, Carolin M. Pirkl⁴, Rolf F. Schulte⁴, Guido Buonincontri¹, Alessandro Arduino⁵, Luca Zilberti⁵, and Michela Tosetti^1,2
1IRCCS Stella Maris, Pisa, Italy, ²Imago7 Foundation, Pisa, Italy, ³Università di Pisa, Pisa, Italy, ⁴GE Healthcare, Munich, Germany, ⁵Istituto Nazionale di Ricerca Metrologica (INRiM), Torino, Italy

MR-EPT aims to non-invasively assess the electrical properties (EP) of the tissues. However, conventional methods based on second-order derivatives of the transmit field are hampered by high noise amplification and boundary errors. Here, we propose to rely on the correlation between electrical properties and water content, which in turn is correlated to tissue T1, in order to extract EP maps from MR Fingerprinting based T1 estimation. We demonstrate whole-brain high-resolution EP mapping of healthy human volunteers and consistent results with both theoretical predictions and standard MR-EPT estimations.

Sebastian Flassbeck^1,2 and Jakob Assländer^1,2
1New York University School of Medicine, Center for Biomedical Imaging, New York, NY, United States, ²Center for Advanced Imaging Innovation and Research, New York University School of Medicine, New York, NY, United States

Low field magnetic resonance systems have potential as powerful diagnostic tools at comparably low cost. However, the inherently lower signal requires the use of SNR-efficient pulse sequences and reconstruction schemes. In this work, we present initial results with a hybrid-state free precession sequence, used to quantify T1 and T2 in 12min with an isotropic resolution of 1mm3. Visually, these quantitative maps show low noise levels, which allows for compelling synthetic MP-RAGE contrast generation.

Andrew Dupuis¹, Yong Chen¹, Rasim Boyacioglu¹, John Stairs², Michael Hansen², Kelvin Chow³, and Mark A Griswold^1
1Case Western Reserve University, Cleveland, OH, United States, ²Microsoft, Redmond, WA, United States, ³Siemens Medical Solutions USA, Inc., Chicago, IL, United States

Reconstruction of 3D Magnetic Resonance Fingerprinting acquisitions is computationally demanding, resulting in long processing times. GPU parallelization of the reconstruction’s NUFFT, pattern matching, and coil combination steps improves performance, but traditionally requires high-performance computers at the scanner. We propose an online reconstruction on a remote GPU-accelerated Kubernetes cluster. This allows many scanners or sites to share easily upgradeable and manageable computing resources. Additional calibration measurements, such as B1 maps, can also be transferred to allow inline B1 inhomogeneity correction. We also demonstrate that 3D-MRF reconstruction is robust with raw data compression that can be used to reduce  site-to-cloud bandwidth requirements.