Tae Hyung Kim^1,2, Jaejin Cho^1,2, Bo Zhao³, and Berkin Bilgic^1,2,4
1Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, ²Radiology, Harvard Medical School, Boston, MA, United States, ³Biomedical Engineering, The University of Texas at Austin, Austin, TX, United States, ⁴Harvard/MIT Health Sciences and Technology, Cambridge, MA, United States

We introduce a novel framework that jointly performs advanced image reconstruction and model-based MR parameter mapping, where various traditional and modern reconstruction techniques and signal relaxation models (T₁, T₂, T₂*, etc) can be integrated as a plug-and-play manner.  Using the proposed framework, we also incorporated model-based parameter mapping with scan-specific deep learning reconstruction (a method named LORAKI).  The experiment results with T₂, T₂* and T₁ indicate that this synergistic combination is advantageous, providing improved quantitative imaging over existing methods, e.g. with up to 3.6-fold, 1.7-fold, 2.3-fold NRMSE gain in T₂, T₂* and T₁  estimation, respectively.

Michael Bach¹, Tom Hilbert^2,3,4, Francesco Santini^5,6, Hanns Christian Breit¹, Markus Klarhöfer^5,7, Davide Piccini^2,3,4, Tobias Kober^2,3,4, and Bénédicte Maréchal^2,3,4
1Department of Radiology, University Hospital Basel, Basel, Switzerland, ²Advanced Clinical Imaging Technology, Siemens Healthcare, Lausanne, Switzerland, ³Department of Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, ⁴LTS5, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland, ⁵Division of Radiological Physics, Department of Radiology, University Hospital Basel, Basel, Switzerland, ⁶Department of Biomedical Engineering, University of Basel, Basel, Switzerland, ⁷Siemens Switzerland AG, Healthcare Sector, Zürich, Switzerland

T1 mapping could potentially accurately diagnose microstructural changes due to pathology. In this work, a mp2rage based approach was optimized to perform fast whole-brain T1 mapping at 0.55 T. The mp2rage method was validated by the inversion recovery (IR) method. Finally, a fully automated brain segmentation was applied to perform region-based analysis of T1 values. Both methods (IR and mp2rage) yield comparable T1 values. The fully automated brain segmentation showed a good performance for the mp2rage sequence at 0.55 T. ROI extraction should further be optimized as it showed outliers most likely due to CSF voxels contaminating the ROIs.

John A Roberts¹, Seong-Eun Kim¹, Evgueni Kholmovski^1,2, Ying Hitchcock³, Tyler John Richards¹, and Yoshimi Anzai^1
1Radiology and Imaging Sciences, University of Utah, Salt Lake City, UT, United States, ²Biomedical Engineering, University of Maryland, Baltimore, MD, United States, ³Radiation Oncology, Huntsman Cancer Institute, Salt Lake City, UT, United States

Arterial input functions measured in T1 weighted low tip angle SPGRE DCE MRI acquisitions may violate an important assumption used when modeling the conversion of image values to quantitative results: that all tissues of interest have reached steady state within the imaging volume.  For arterial flow along the short axis of the imaging volume, blood may not dwell within the imaging volume long enough to reach steady state.  We study both the location from which AIF will be measured and the acquisition parameters employed in order to optimize the accuracy of the DCE derived quantitative results.

Phillip DiGiacomo¹, Wei Shao¹, Marios Georgiadis¹, Yi Wang², Pascal Spincemaille², Philipp Schlömer³, Markus Axer³, Don Born¹, Mirabela Rusu¹, and Michael Zeineh^1
1Stanford University, Stanford, CA, United States, ²Cornell University, Ithaca, NY, United States, ³Research Centre Jülich, Jülich, Germany

Histology is critical for validating pathology detected in ultra-high-resolution ex-vivo MRI. However, registering histological images with MR is challenging due to 3D deformations between MR and histology, 2D deformations during sectioning, and the fact that histological staining often occurs only in selected slices. Addressing these issues will enable quantification of complex histological features and their relationship with MRI, facilitating the discovery of novel in vivo imaging-based biomarkers of disease. Here, we demonstrate the use of a novel pipeline utilizing ultra-high-resolution specimen MRI, blockface imaging, and a modified tape transfer system for correlative MRI-paraffin-embedded histology of human brain tissue specimens.

Jiachen Zhuo¹, Muhan Shao², Steven Roys¹, Xiao Liang¹, Rosy Linda Njonkou Tchoquessi¹, Prashant Raghavan¹, Jerry Prince², and Rao Gullapalli^1
1Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, Baltimore, MD, United States, ²Electrical and Computer Engineering, Johns Hopkins University, Baltimore, MD, United States

The perivascular space (PVS) is key to brain waste clearance and brain metabolic homeostasis. Enlarged PVS (ePVS) can be automatically quantified reliably by combining the 3D T1w and 3D T2w images to produce enhanced PVS contrast followed with frangi filtering and thresholding. However often times only 2D T2w images are available, especially in clinical exams. In this study, we investigate the feasibility of using an innovative deep learning based super-resolution technique (SMORE) to produce 3D T2w images (SR T2) for ePVS quantification. We show that the SR T2 volume provided comparable ePVS estimation as a 3D T2 volume.

Brian Johnson^1,2, John Penatzer¹, and Sandeep Ganji^1,3
1Philips Healthcare, Gainesville, FL, United States, ²University of Texas Southwestern Medical Center, Dallas, TX, United States, ³Mayo Clinic College of Medicine, Rochester, MN, United States

Simultaneous parametric mapping techniques like 3D-QALAS offer means for reliable measurement of T1, T2, and proton density while also offering the generation of standard imaging contrasts.  3D-QALAS has also been shown to provide consistent brain tissue segmentation and volumetric analysis.  However, high-resolution 3D-QALAS scans suffer from long scan times.  Applying acceleration techniques like compressed SENSE, which are highly effective at reducing 3D scan times, can bring 3D-QALAS scan times under 3-minutes.  Here we evaluate the acceleration of 3D-QALAS using multiple compressed SENSE factors on scan times, image quality, multiparametric mapping, and volumetric analysis.

Gyula Kotek¹, Willem van Valenberg¹, Laura Nunez-Gonzalez¹, Dirk H. J. Poot¹, Mika W. Vogel², and Juan A. Hernandez-Tamames^1
1Radiology and Nuclear Medicine, Erasmus MC, Rotterdam, Netherlands, ²GE Healthcare, Hoevelaken, Netherlands

Unlike multi-parametric methods as MR Fingerprinting and quantitative transient imaging we present a quantitative MRI method based on the transient response without variation of experimental parameters (FA…). Our method is based on the exact solution of the recursive equation for the magnetization along an acquisition train. Based on this analytical approach we determine the requirements for the features and parameters of a repeated acquisition block to allow simultaneous estimation of experimental (B0, B1) and intrinsic (PD, T1, T2) parameters. The feasibility of the technique on clinical scanners is demonstrated by acquiring a comprehensive set of 3D in vivo parametric maps.

Hao Li¹, Charlotte E Buchanan², David M Morris³, Alexander J Daniel², João Sousa⁴, Steven Sourbron⁴, David L Thomas^5,6,7, Susan T Francis², and Andrew Nicholas Priest^1,8
1Department of Radiology, University of Cambridge, Cambridge, United Kingdom, ²Sir Peter Mansfield Imaging Centre, University of Nottingham, Nottingham, United Kingdom, ³Centre for Cardiovascular Science, University of Edinburgh, Edinburgh, United Kingdom, ⁴Department of Infection, Immunity and Cardiovascular Disease, University of Sheffield, Sheffield, United Kingdom, ⁵Neuroradiological Academic Unit, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom, ⁶Dementia Research Centre, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom, ⁷Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom, ⁸Department of Radiology, Addenbrooke’s Hospital, Cambridge, United Kingdom

B1 inhomogeneity and non-ideal slice profiles introduce contributions from stimulated and indirect echoes into the multi-echo spin-echo (MESE) T₂ mapping sequence, leading to variations in quantitative T₂ values across scanners and vendors despite the use of a harmonised scan protocol. This study used an EPG-based fitting method to include corrections for stimulated-echo effects, applied to phantom and ‘travelling kidney’ data collected on scanners from three different MR vendors (GE, Philips, Siemens). Compared with conventional monoexponential fitting, EPG-based fitting substantially reduced the inter-scanner variations of T₂ measurements. This improves the harmonization of the MESE T₂ mapping sequence across MR scanner vendors.

Samantha Paterson^1,2, Antoine Vallatos³, Camille Graff⁴, and William Matthew Holmes^2
1EPIC, University of Leeds, Leeds, United Kingdom, ²GEMRIC, University of Glasgow, Glasgow, United Kingdom, ³Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, United Kingdom, ⁴INP Grenoble - Phelma, Grenoble, France

The mbASL sequence is a hybrid between PASL and pCASL. It was shown to produce high SNR perfusion data. Quantification is crucial in order to produce accurate mbASL CBF maps. We have shown successful quantification by modifying the Buxton ASL kinetic model, optimising the number of pulses used for inversion and examining the optimal labelling slab thickness needed to maximise the SNR. We found CBF values that agree with current literature for mice and rats.

Hernán Arturo Mella^1,2,3 and Sergio Uribe^2,3,4
1Department of Electrical Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile, ²Biomedical Imaging Center, Pontificia Universidad Católica de Chile, Santiago, Chile, ³ANID - Millennium Science Initiative Program - Millennium Nucleus in Cardiovascular Magnetic Resonance, Santiago, Chile, ⁴Department of Radiology, School of Medicine, Pontificia Universidad Católica de Chile, Santiago, Chile

In this work, we investigated the influence of fat in the estimation of strain from tagged MR images. The results showed that in a static phantom with sharp and smooth fat fractions, the estimated strain is influenced by the bandwidth of the bandpass filter, the tagging period, and the fat fraction.

Anthony Tessier¹, Anthony Ruze¹, Emilien Royer¹, Monique Bernard¹, Angele Viola¹, and Teodora-Adriana Perles-Barbacaru^1
1CRMBM UMR 7339, Aix Marseille University, CNRS, Marseille, France

Maps of the contrast agent concentration over time in mouse brains upon intraperitoneal administration were obtained by a dynamic MRI technique. The mice had different degrees of contrast agent accumulation clearly distinguishable on the maps with a particular distribution resembling the pathways of the glymphatic system. The average brain concentration computed 2 hours post contrast agent administration correlates with the gadolinium dosage in the brain by inductively coupled plasma mass spectroscopy proving that quantification is feasible although the signal analysis can be improved. 

Constanza Gaínza^1,2, Daniele E Schiavazzi³, and Carlos A Sing-Long^1,2,4,5,6
1Institute for Mathematical and Computational Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile, ²Millennium Nucleus Center for the Discovery of Structure in Complex Data, Santiago, Chile, ³Department of Applied and Computational Mathematics and Statistics, University of Notre Dame, Notre Dame, IN, United States, ⁴Institute for Biological and Medical Engineering, Pontificia Universidad Católica de Chile, Santiago, Chile, ⁵Millennium Nucleus for Cardiovascular Magnetic Resonance, Santiago, Chile, ⁶Biomedical Imaging Center, Pontificia Universidad Católica de Chile, Santiago, Chile

The wall shear stress (WSS) is a relevant biomarker associated with the rupture of atherosclerotic plaques that can be computed from blood flow velocity measurements. We study the effect of noise velocity on the quantification of the WSS with a focus on the spatial correlations that may arise. We perform in silico experiments in which we consider a Hagen-Poiseuille flow and two noise distributions. Our results show evidence on the existence of spatial correlations in the WSS even when the velocity noise is uncorrelated.

Martin Soellradl¹, Christina Graf¹, and Rudolf Stollberger^1
1Institute of Medical Engineering, Graz University of Technology, Graz, Austria, Graz, Austria

To exploit acceleration potential of Bloch-Siegert shift based $$$B_1^+$$$-mapping, we implemented a 3D segmented EPI sequence with variable number of interleaves. Additionally, sampling of the k-space center with variable block-sizes was investigated. Compared to a fully sampled gradient-echo sequence, we show that acceleration by a factor up to 50 is feasible. Depending on the desired accuracy, whole-brain 3D $$$B_1^+$$$-mapping is possible in 5 to 10 seconds.