PINAR SENAY ÖZBAY^1,2, Cristina Rossi¹, Geoffrey Warnock³, Klaas Paul Prüssmann², and Daniel Nanz^1
1Institute for Diagnostic and Interventional Radiology, University Hospital Zürich, Zürich, Zürich, Switzerland, ²Institute for Biomedical Engineering, University and ETH Zürich, Zürich, Zürich, Switzerland, ³University Hospital Zürich, Clinic of Nuclear Medicine, Zürich, Zürich, Switzerland

In functional MRI, phase information as complementary information to BOLD signal-magnitude changes have been investigated with increasing interest in recent years. In this study we tested to what degree Quantitative Susceptibility Maps, derived from EPI-phase data that were simultaneously acquired with traditional BOLD fMRI magnitude data, reflect patterns of neuronal activation as observed with an identical block-design visual-stimulation paradigm in water PET and BOLD MRI data. Observations pertinent to data processing and QSM data are reported.

PINAR SENAY ÖZBAY^1,2, Michael Wyss², Cristina Rossi¹, Klaas Paul Prüssmann², and Daniel Nanz^1
1University Hospital Zürich, Institute for Diagnostic and Interventional Radiology, Zürich, Zürich, Switzerland, ²Institute for Biomedical Engineering, University and ETH Zürich, Zürich, Zürich, Switzerland

In-vivo imaging of deep cerebellar structures and deep nuclei is difficult even by MRI due to their small size and limited contrast. In this work we aimed at a visualization and characterization of the fine corrugated band that represents the dentate nucleus (DN) by means of high-resolution Quantitative Susceptibility Mapping (QSM) at 7T. QSM depicted the DN band with significantly better contrast than T2*-weighted images or T2* maps. Quantitative susceptibility differences of the right and left DN band with respect to frontal white matter could be measured. QSM-based estimates of right and left DN band volumes are also reported.

Matthew Cronin¹, Samuel Wharton¹, Penny Gowland¹, and Richard Bowtell^1
1Sir Peter Mansfield Magnetic Resonance Centre, The University of Nottingham, Nottingham, Notts, United Kingdom

The magnetic field perturbations produced when simple structures composed of materials with isotropic magnetic susceptibility are exposed to magnetic fields have been well documented, but the effect of material of anisotropic magnetic susceptibility has been less explored. Here, we experimentally investigate the field perturbations produced by spherical shells of pyrolytic graphite sheet, a material with a highly anisotropic susceptibility. It is shown that in agreement with theory, uniform, geometry-dependent field offsets are produced inside the spheres with minimal external field perturbation. The results indicate that hollow spheres with anisotropic susceptibility could form the basis of useful tunable contrast agents.

Way Cherng Chen^1,2, Karla Miller², and Kai-Hsiang Chuang^1
1Singapore Bioimaging Consortium, Singapore, Singapore, Singapore, ²FMRIB, University of Oxford, Oxford, Oxon, United Kingdom

The relationship between white matter microstructure and GRE signal phase and magnitude was investigated using a cuprizone mouse model of demyelination. Ex vivo R2* and frequency maps showed high correlations with histological myelin stain intensity. Geometric WM modeling suggested that the GRE signal is not only sensitive to the bulk amount of myelin but also the spatial distribution of myelin within WM.

Ferdinand Schweser¹, Edsel Daniel Peres Gomez¹, Andreas Deistung¹, and Jürgen R Reichenbach^1
1Medical Physics Group, Institute of Diagnostic and Interventional Radiology I, Jena University Hospital - Friedrich Schiller University Jena, Jena, Germany

In a recent paper Liu and Li presented a novel MR phase-based technique to assess tissue anisotropy, in the following referred to as Fourier spectrum Tensor Imaging (FTI). With this contribution we provide important insights on the most important parameters of FTI: the MR imaging resolution, the spatial resolution of the numerical grid during the post-processing, the effect of the size of the ROI, and the sensitivity to measurement noise.

Matthew Cronin¹, Samuel Wharton¹, Anna Blazejewska¹, Penny Gowland¹, and Richard Bowtell^1
1Sir Peter Mansfield Magnetic Resonance Centre, The University of Nottingham, Nottingham, Notts, United Kingdom

As gradient echo phase images are increasingly being used in MRI, it is important to understand factors which contribute to MR signal phase. Here, using numerical simulations, we investigate the effect of oriented, spheroidal, NMR-invisible perturbers of varying volume fraction and shape on the average frequency of the NMR signal for different diffusion rates. Results indicate that the frequency is generally dependent on the perturber shape and does not simply reflect the volume average susceptibility. The frequency is simply related to the perturber’s demagnetising factor at short evolution times and at all times when the rate of diffusion is high.

Christian H Ziener^1
1Radiology, German Cancer Research Center, Heidelberg, Baden-Württemberg, Germany

Signal formation in the dipole field of a single vessel is of paramount interest in susceptibility weighted imaging and functional MRI. Spin dephasing around a vessel is analyzed with consideration of diffusion effects which cannot be neglected for small vessels. A full analytic solution of the Bloch-Torrey equation is presented without any approximations, thus enabling the correct interpretation of the susceptibility weighted signal. With these solutions at hand, SWI and DWI measurements can be interpreted at the same time, since the influence of diffusion on magnitude and phase images is included in this analysis.

Meng-Chi Hsieh^1,2, Chung-Ming Chen³, Kun-Hsien Chou⁴, Ai-Ling Hsu^1,2, Ching-Po Lin⁴, and Jyh-Horng Chen^1,2
1Graduate Institute of Biomedical Electronics and Bioinformatics, National Taiwan University, Taipei, Taiwan, ²Interdisciplinary MRI/MRS Laboratory, Department of Electrical Engineering, National Taiwan University, Taipei, Taiwan, ³Institute of Biomedical Engineering, National Taiwan University, Taipei, Taiwan, ⁴Brain Connectivity Lab, Institute of Neuroscience, National Yang-Ming University, Taipei, Taiwan

The purpose of this study is to improve accuracy for quantitative susceptibility map (QSM) by Edge-Weighted L1 (EW-L1) regularization. Numerical simulation and human brain experiment were performed to demonstrate the superiority of EW-L1 over conventional L1 method. Our results showed that EW-L1 provides highly accurate estimation and suppressed streaking artefact of QSM. Preliminary results of brain imaging suggested its feasibility of QSM in clinical practice.

Wen-Tung Wang¹, Ningzhi Li², Dzung Pham³, and John Anthony Butman^4
11Center for Neuroscience and Regenerative Medicine, Henry Jackson Foundation, Bethesda, MD, United States, ²Center for Neuroscience and Regenerative Medicine, Henry Jackson Foundation, MD, United States, ³1Center for Neuroscience and Regenerative Medicine, Henry Jackson Foundation, MD, United States, ⁴Diagnostic Radiology, National Institute of Health, MD, United States

A number of phase unwrapping algorithms can faithfully unwarp wrapped MRI phase images on a voxel-by-voxel basis. These sequential algorithms cannot unwrap phase fringelines, which terminate at singular poles. The method based on Fourier properties and trigonometric identities of the Laplacian operator can unwrap wrapped raw phase images with fringelines, but its resultant images show low-frequency deviation. By combining the Fourier method with a sequential algorithm by Cusack et al. using high- and low-pass filtering, the wrapped raw phase with fringelines can be unwrapped without low-frequency deviations.

Ryota Sato¹, Toru Shirai¹, Yo Taniguchi¹, Takenori Murase², Yoshihisa Soutome^1,2, Yoshitaka Bito², and Hisaaki Ochi^1
1Central Research Laboratory, Hitachi, Ltd., Kokubunji, Tokyo, Japan, ²Hitachi Medical Corporation, Chiba, Japan

To achieve a susceptibility map with high accuracy and high precision, a method is proposed for QSM with a combination of different regularization parameters. In the proposed method, k-space was divided into three domains and susceptibility maps calculated by different regularization parameters are applied to each domain. A numerical simulation based on a COSMOS reconstructed susceptibility map was used to compare both the accuracy and the precision of susceptibility maps calculated by conventional and proposed methods. The results suggest that the accuracy and the precision of the proposed method are comparable to or higher than those of the conventional methods.

Saifeng Liu¹, Sagar Buch¹, and E. Mark Haacke^1,2
1School of Biomedical Engineering, McMaster University, Hamilton, Ontario, Canada, ²Department of Radiology, Wayne State University, Detroit, Michigan, United States

The Laplacian of phase has been frequently used in background phase removal in QSM. To calculate the Laplacian of phase, the wrapped phase differences together with discrete Laplacian operator can be used, or the trigonometric functions together with Fourier transform (continuous Laplacian operator) can be used. Former studies have shown errors associated with veins in the phase images processed using continuous Laplacian operators. In this study, both discrete and continuous Laplacian operators are evaluated using simulated and in vivo data, in order to understand the source of the error associated with veins in the processed phase images.

Job G Bouwman¹ and P R Seevinck^2
1Image Sciences Institute, University Medical Center, Utrecht, Utrecht, Netherlands, ²Image Sciences Institute, University Medical Center, Utrecht, Netherlands

Experimental validation of how QSM reconstruction can be improved by preconditioning the input field, based on a prior estimate of the underlying susceptibility distribution. Using prior knowledge, a better distinction was made between background and foreground field leading to a more valid dχ-reconstruction. Moreover, by flattening the input field, less regularization was needed, leading to sharper results, less streaking and reduced underestimation.

Yuya Umemoto¹, Mai Murashima^2,3, Tomohiro Ueno², and Naozo Sugimoto^2
1Faculty of Medicine, Kyoto University, Kyoto, Kyoto, Japan, ²Graduate of Medicine, Kyoto University, Kyoto, Kyoto, Japan, ³present address Toshiba Medical Systems Corporation, Otawara, Tochigi, Japan

In Quantitative Susceptibility Mapping, susceptibility distribution can be obtained by deconvolving a perturbed magnetic field with a spatial unit dipole field. The spatial dipole field has a rapid changing nature. In this study, we employed a denser sampled dipole field in the deconvolving process to improve spatial resolution of QSM. We created Shepp-Logan phantom including partial volume effects and used as a low resolution input data. A Fourier-transformed dipole field with high resolution was created analytically, and a perturbed field was expanded by the nearest neighbor method. With the proposed method, a susceptibility map with higher spatial resolution was obtained.

Saifeng Liu¹, Sagar Buch¹, and E. Mark Haacke^1
1School of Biomedical Engineering, McMaster University, Hamilton, Ontario, Canada

The accuracy of quantitative susceptibility mapping is dependent on the background phase removal. When regions with unreliable phase values are included in the phase image, e.g. vessels without proper flow compensation or bones with noisy phase, a simple background phase removal may amplify phase noise or create edge artifacts. In this study, we show a method to reduce the phase artifacts that are caused by the inclusion of noisy regions with unreliable phase values, in order to improve the accuracy of QSM.

Johannes Lindemeyer¹, Michael S. Poole¹, Desmond H. Y. Tse¹, Ana-Maria Oros-Peusquens¹, and N. Jon Shah^1,2
1INM - 4, Research Centre Jülich GmbH, Jülich, Germany, ²Department of Neurology, RWTH Aachen University, Aachen, Germany

Gradient echo based phase mapping for susceptibility reconstructions in the human brain requires careful data evaluation at ultra-high field. We present a workflow including recombination of multiple receive channels and advanced processing methods for obtaining background corrected field and susceptibility maps. Correction of the individual receive channels for spatially varying phase offsets represents an essential processing step.

Chun-Kun Wang¹, Po-Yu Lin², Tzu-Cheng Chao^1,2, and Ming-Long Wu^1,2
1Institute of Medical Informatics, National Cheng Kung University, Tainan, Taiwan, ²Department of Computer Science and Information Engineering, National Cheng Kung University, Tainan, Taiwan

Quantitative susceptibility mapping has been a useful tool to monitoring magnetic properties of the tissues. However streaking artifacts of most QSM reconstruction hampers the interpretation of the susceptibility maps. While several reconstruction algorithms were proposed to alleviate the artifact, an object quantification of artifact is also important. The presented work aims at developing an automatic detection of the residual streaking artifact to facilitate an objective assessment of quality of QSM reconstructions.

Ryota Sato¹, Toru Shirai¹, Yo Taniguchi¹, Takenori Murase², and Hisaaki Ochi^1
1Hitachi, Ltd., Kokubunji, Tokyo, Japan, ²Hitachi Medical Corporation, Chiba, Japan

To achieve SWI based on susceptibility maps without streaking artifacts in short calculation time, a method was developed for susceptibility-weighted imaging using a susceptibility map calculated without regularization in very small iteration number. To evaluate the usefulness of the proposed method, image contrasts of axial slices in closed-type MRI and open-type MRI acquired by the conventional and the proposed method were compared. The results suggest that the proposed method can enhance contrast of regions with high magnetic susceptibility, such as veins or iron depositions, in arbitrary slice orientation within about 20 seconds without streaking artifacts.

Guillaume Gilbert^1,2 and Laurent Létourneau-Guillon^2
1MR Clinical Science, Philips Healthcare, Montreal, Quebec, Canada, ²Department of Radiology, Centre Hospitalier de l'Université de Montréal, Montreal, Quebec, Canada

The purpose of this work was to evaluate the use of 3D multi-echo flow compensation for improved multi-echo susceptibility-weighted imaging, to extend a concept for partial flow compensation to multi-echo imaging and to perform a direct comparison between complete and partial flow compensation methods. The inclusion of 3D multi-echo flow compensation is shown to improve arterial depiction in the combined susceptibility-weighted images. Partial flow compensation is shown to provide substantial improvements while not requiring as much of a compromise in terms of echo spacing as complete flow compensation.

Kofi Deh¹, Sarah Eskreis-Winkler², Pascal Spincemaille², Thanh Nguyen², and Yi Wang^1,3
1Weill Cornell Medical College, New York, NY, United States, ²Weill Cornell Medical College, New York, United States, ³Department of Biomedical Engineering, Cornell University, NY, United States

The amount of variation between quantitative susceptibility maps (QSM) of the same subject acquired with different MRI systems has remained an open question. We investigated this problem by comparing QSM images acquired at two different sites, one with a Siemens 3T and the other with a GE 3T scanner. We obtained high correlation coefficients for QSM images from the two sites (0.99 for phantoms, 0.97 for humans). Taken together with our Bland-Altman plots of the data, which indicated no bias, we conclude that scanner systems we studied may be used interchangeably.

Jie Luo¹ and Dmitriy Yablonskiy^1
1Radiology, Washington University in St. Louis, St. Louis, MO, United States

The phase contrasts obtained by gradient echo MRI provides complimentary information on brain anatomy. One of the ways to understanding the biophysical origins of the phase contrast relies on postmortem studies. However, to what extend and how fast postmortem changes affect MR tissue properties are not known. In this pilot study, we have demonstrated for the first time changes in the gray/white matter phase contrast and R2* during transition from live to postmortem conditions. Our data are in agreement with minor role of blood oxygenation level on gray/white matter phase contrast and major role on changes in R2* contrast.

Li Huang¹, Ferdinand Schweser², Karl-Heinz Herrmann², Martin Krämer², Andreas Deistung², and Jürgen Rainer Reichenbach^2
1in-vivo-MR Group, Faculty 02 (Biology/Chemistry), University Bremen, Bremen, Bremen, Germany, ²Medical Physics Group, Institute of Diagnostic and Interventional Radiology I, Jena University Hospital - Friedrich Schiller University Jena, Jena, Thuringia, Germany

MRI-based conductivity mapping is a promising new MR technique with great potential for clinical applications. Literature algorithms for reconstructing conductivity maps suffer from edge artifacts near tissue interfaces, which restrict the reliability of the conductivity information. To overcome this limitation we propose a novel Monte Carlo method, which was applied in a dedicated phantom and in vivo experiment. Resulting conductivity maps showed improved SNR with well delineated anatomic structures, reduced edge artifacts, and reasonable conductivity values.

Andrea Borsic¹, Irina Perreard², and Ryan J Halter^1
1Thayer School of Engineering, Dartmouth College, Hanover, NH, United States, ²Radiology Department, Dartmouth College, Hanover, NH, United States

This manuscript presents a novel approach to the reconstruction of the electrical properties of tissues from MRI B1 field information. While other approaches require differentiation of B1 phase or amplitude information, an inverse problem approach is proposed, where conductivity / permittivity are estimated by matching a simulated B1 phase / amplitude to measured ones. Through this formulation it is possible to reduce sensitivity to noise typical of differentiation processes and to deal better with sharp image transitions.

Eric Michel¹, Daniel Hernandez¹, Min Hyoung Cho¹, and Soo Yeol Lee^1
1Dept. of Biomedical Engineering, Kyung Hee University, Yongin-Si, Gyeonggi-Do, Korea

Computing electrical conductivity or electrical permittivity maps from the measured B1 maps is a noise-amplifying process due to the Laplacian computation involved in solving the underlying electromagnetic field equations. We applied adaptive diffusion filters (ADFs) to denoise the B1 maps prior to the ordinary Laplacian computation based on the second-order difference equation on three consecutive neighboring pixels in each direction. Both in the simulations and in-vivo conductivity imaging experiments at 3T, we observed significant improvements in the conductivity image quality computed from the noisy B1 maps acquired by the double angle method (DAM).

Selaka B Bulumulla¹, Jeannette C Roberts¹, Seung-Kyun Lee¹, Peter Lamb¹, and Ileana Hancu^1
1GE Global Research, Niskayuna, NY, United States

Dielectric contrast of tumors has received significant interest as a potential means to improve specificity of cancer imaging. Although dielectric property estimation from MR transmit field is a promising approach, the noise inherent in field maps and the reconstruction algorithm may introduce errors. Therefore, we propose a reference aided method to validate data sets, and optimize reconstruction algorithm parameters to improve the process. The process is demonstrated with two reference solutions and adenocarcinoma strains grown in rat models.

Jiaen Liu¹, Xiaotong Zhang¹, Pierre-Francois Van de Moortele², Sebastian Schmitter², and Bin He^1,3
1Biomedical Engineering, University of Minnesota, Minneapolis, Minnesota, United States, ²Center for Magnetic Resonance Research, University of Minnesota, Minneapolis, Minnesota, United States, ³Institute for Engineering in Medicine, University of Minnesota, Minneapolis, Minnesota, United States

Electrical properties tomography (EPT), which utilizes measurable radiofrequency field (B1) information to image the electrical properties (EP) of tissue, including conductivity and permittivity, holds promises in both clinical diagnosis and real time subject-specific quantification of specific absorption rate. In a previous study, we have derived and validated a gradient-based EPT (gEPT) approach with significantly improved boundary reconstruction and fidelity against measurement noise. Here, a more generalized method for gEPT is introduced to quantitatively solve gEPT under the practical situation when proton density is unknown. As a result, both EP and proton density can be obtained with satisfactory boundary reconstruction.