Electro-Magnetic Tissue Properties Mapping

Monday 1 June 2015

Woo Chul Jeong¹, Saurav ZK Sajib¹, Ji Eun Kim¹, Hyung Joong Kim¹, Oh In Kwon², and Eung Je Woo^1
1Kyung Hee University, Yongin, Gyeonggi, Korea, ²Konkuk University, Seoul, Korea

Estimation of ablated lesion and control of RF power is important to reduce local recurrence after RF ablation, there still exist needs for new bio-imaging technique providing a non-invasive monitoring of RF ablation. We proposed MR-based fast conductivity imaging for monitoring of RF ablation using biological tissues. Fast MREIT produces conductivity images at every 10.24 seconds during RF ablation. We could extract spatiotemporal maps of tissue conductivity from a time series of 180 conductivity images for 30.72 minutes. Analyzing the time series of conductivity images, we could distinguish six different stages and separately interpret temperature-dependent and/or structure-dependent conductivity changes.

Woo Chul Jeong¹, Min Oh Kim², Saurav ZK Sajib¹, Ji Eun Kim¹, Hyung Joong Kim¹, Oh In Kwon³, Dong Hyun Kim², and Eung Je Woo^1
1Kyung Hee University, Yongin, Gyeonggi, Korea, ²Yonsei University, Seoul, Korea, ³Konkuk University, Seoul, Korea

Electrical tissue conductivity is primarily determined by concentration and mobility of ions in intra- and extra-cellular fluids. The conductivity of biological tissues show frequency-dependent behavior and its values at different frequencies may provide useful diagnostic information. MR-based tissue property mapping are widely used imaging techniques which provided unique conductivity contrast at different frequency ranges. Recently, new method for data acquisition and reconstruction for low- and high-frequency conductivity images from a single MR scan was proposed. In this study, we applied the simultaneous dual-frequency range conductivity mapping to enhance its clinical potentials through in vivo disease model animal imaging.

Sung-Min Gho¹, Jaewook Shin¹, Min-Oh Kim¹, Dongyeob Han¹, and Dong-Hyun Kim^1
1Electrical and Electronic Engineering, Yonsei University, Sinchon-dong, Seoul, Korea

MR imaging can provide various quantitative information regarding the electro-magnetic properties and relaxation properties of tissue. The electric and magnetic properties are linked through Maxwell's equations and the magnetic susceptibility is one major source of the R2* and R2'. These quantitative information can be used independently, however, studies which use several related quantitative information together can be useful since mis-registration, different physiological noise and lengthened scan time can be alleviated due to separate measurements. In this abstract, we propose a new method for obtaining these quantitative applications simultaneously. Therefore, quantitative conductivity map, QSM, R2*, and R2' maps.

Saurav ZK Sajib¹, Ji Eun Kim¹, Woo Chul Jeong¹, Hyung Joong Kim¹, Oh In Kwon², and Eung Je Woo^1
1Kyung Hee University, Yongin, Gyeonggi, Korea, ²Konkuk University, Seoul, Korea

Electrical tissue conductivity is primarily determined by concentration and mobility of ions. If electrical current is carried by dissolved ions, the more ions generate more currents. In addition, the mobility of ions is restricted by heterogeneous membrane, current density of tissue dependent on the physical environment even though it has uniform ion concentration. Using MREIT technique, we provide electrical conductivity-based metabolite mapping with highly sensitive to its concentration changes. Together with in vitro measurement of metabolites by concentration changes, we performed phantom imaging to evaluate the proposed method showing a significant relationship between concentration and electrical conductivity of brain metabolites.

Hyunyeol Lee¹ and Jaeseok Park^2
1Center for Neuroscience Imaging Research, Institute for Basic Science (IBS), Sungkyunkwan University, Suwon, Gyeonggi, Korea, ²Department of Global Biomedical Engineering, Sungkyunkwan University, Suwon, Gyeonggi, Korea

Conventional MREIT suffers from long imaging time and low phase sensitivity, potentially resulting in inefficient conductivity estimation. Current-controlled alternating SSFP-FID was proposed as an alternative to achieve rapid encoding as well as high nonlinear phase sensitivity. Nevertheless, since the nonlinear phase model includes both tissue relaxation and Bz properties, the former has to be known as a priori for conductivity estimation, which requires multiple separate acquisitions. In this work, we propose a novel, current-controlled alternating reversed dual echo steady state (DESS) MREIT for joint estimation of tissue relaxation and electrical properties in a single measurement.

S. Mandija¹, A.L.H.M.W. van Lier¹, P. Petrov², S.W.F. Neggers², P.R. Luijten¹, and C.A.T. van den Berg^1
1Imaging Division, UMC Utrecht, Utrecht, Netherlands, ²Brain Center Rudolf Magnus, UMC Utrecht, Utrecht, Netherlands

In this abstract we demonstrate that combinations of phase maps obtained with different gradient polarities should be done carefully since these images are prone to geometrical shift leading to RF phase falsely attributed to LF phase. Since the RF phase is related to conductivity at MHz frequency, because of this geometrical displacement it appears that the LF phase is also proportional to electrical conductivity. Instead, as shown in simulations, if corrections are performed, the LF phase is not measurable anymore, thus the scaling of the LF phase with the conductivity can be attributed to the RF phase leakage.

Eric Duggan Gibbs^1,2 and Chunlei Liu^2,3
1Biomedical Engineering, Duke University, Durham, NC, United States, ²Duke University Medical Center, Brain Imaging and Analysis Center, Durham, NC, United States, ³Department of Radiology, Duke University, Durham, NC, United States

Eddy-currents generated by rapidly varying gradients have been proposed as a current source to map electrical properties of tissue non-invasively at low frequencies. It has been suggested that charge accumulated at electrical property interfaces potentially impacts the recorded MR signal. Induced charges also may impact images with conductive prosthetics. This work outlines and demonstrates an algorithm that determines electromagnetic fields in non-homogeneous materials and accounts for charge accumulation due to eddy-currents induced by rapidly varying gradients.

Jiasheng Su¹, Bingwen Zheng², Sam Fong Yau Li², and Shao Ying Huang^1
1Singapore University of Technology and Design, Singapore, Singapore, ²Department of Chemistry, National University of Singapore, Singapore

The method of retrieving conductivities of human tissues through eddy currents induced by pulsed field gradient was proposed and has been studied intensively. In this abstract, an optimized electromagnetic (EM) model is proposed to study the relation among encoding gradient fields, the induced eddy currents, and the resultant phase difference that was used for the retrieval of conductivity. It is further applied to study the relationship of the gradient fields, the pulse sequence, and the resultant phase difference. Besides the discharging process when the gradient field is a constant that was identified previously, another crucial discharging process to generate phase difference is identified. This discharging process offset the phase difference generated by that when gradient field is constant. It is terminated when an RF pulse is applied. This process is critical for obtaining meaningful phase maps for the retrievals of conductivity. Both theoretical and experimental results are presented.

Eric Michel¹, Daniel Hernandez¹, Min Hyoung Cho¹, and Soo Yeol Lee^1
1Kyung Hee University, Suwon, Gyeonggi-Do, Korea

Measuring the electrical properties of tissue non-invasively with high resolution and accuracy has practical significance for several diagnostic and therapeutic applications in the biomedical field. In this work, the relationship between the electrical properties of tissue and their amount of electrolyte content is exploited. This model is a combination of previous observations of mixture theory and experimental measurements given by literature. We validate our formulation by performing in-vivo EPT mapping of brain tissue at 3T MRI. Great accuracy with not precedent resolution was achieved and we believe this method can have a great impact in EPT estimations.

Edmond Balidemaj¹, Peter de Boer¹, Hans Crezee¹, Rob Remis², Lukas Stalpers¹, Aart Nederveen³, and Cornelis A.T. van den Berg^4
1Radiotherapy, Academic Medical Center, Amsterdam, Netherlands, ²Circuits and Systems Group, TU Delft, Delft, Netherlands, ³Radiology, Academic Medical Center, Amsterdam, Netherlands, ⁴Radiotherapy, UMC Utrecht, Utrecht, Netherlands

In this work we present in vivo reconstructed conductivity values of cervical cancer patients using Electric Properties Tomography (EPT) based on B1+ at 3T. Conductivity values of muscle, bladder and cervical tumors are presented. The results demonstrate the importance of accounting for conductivity values in living conditions when incorporating electric properties data into numerical models.

Jan Sedlacik¹, Ulrich Katscher², and Jens Fiehler^1
1University Medical Center Hamburg-Eppendorf, Hamburg, Hamburg, Germany, ²Philips Research Europe, Hamburg, Germany

MR-based Electric Properties Tomography (EPT) provides a non-invasive means to assess electric tissue properties such as conductivity. EPT was shown to depend on ion concentration and temperature. In this study the effect of ion size, and therefore ion mobility, was studied. The EPT-conductivity steeply changed between HCl and NaOH which is caused by the much smaller free protons of HCl as compared to NaOH. On the other side, the contribution of the much larger glutamate ion on EPT-conductivity is negligible, effectively bisecting the amount of EPT-detectable ions and resulting in nearly half conductivity of MSG as compared to NaCl.

Kathleen M Ropella¹ and Douglas C Noll^1
1Biomedical Engineering, University of Michigan, Ann Arbor, MI, United States

Phase-based conductivity mapping makes use of a noise amplifying operation and often relies on spatial filtering to reduce noise. This work describes a regularized, model-based approach to conductivity mapping, which is more robust in the presence of noisy phase maps and provides better reconstruction near boundaries. We demonstrate the efficacy of the algorithm in simulations as well as in the human brain at 3.0T. This method provides higher SNR and lower RMSE for reconstructed conductivity maps as compared to a basic spatial filtering approach.

Seung-Kyun Lee¹, Selaka Bandara Bulumulla¹, and Ileana Hancu^1
1GE Global Research, Niskayuna, NY, United States

We present calculation of the random noise-limited signal-to-noise ratio (SNR) in MR-based electrical properties tomography (MREPT). We find that the SNR in the relative permittivity and electrical conductivity is determined primarily by the SNR of the magnitude and the phase of the B1+ map, respectively. In addition, the SNR is proportional to the square of the linear dimension of the region-of-interest, to the square root of the number of voxels, and to the inverse square of the RF length scales in the medium. Our results can inform design of MREPT experiments with a desired SNR.

Jiaen Liu¹, Xiaotong Zhang¹, Yicun Wang¹, Pierre-Francois Van de Moortele², 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

The in vivo electrical conductivity and permittivity of tissues at Larmor frequency carries important information for both diagnostic purpose and real-time subject-specific local SAR quantification. Traditional electrical properties tomography (EPT) derives a local solution and can be severely deteriorated due to noise contamination. In the present study, a new approach was proposed to improve the local EPT solution with global regularization by the spatial gradient information of electrical properties, which is deducted from measured transmit B1 field induced in a multi-channel radiofrequency coil. Experiments involving a controlled physical phantom and healthy human subject were conducted to evaluate the proposed algorithm.

Necip Gurler¹, Omer Faruk Oran¹, and Yusuf Ziya Ider^1
1Department of Electrical and Electronics Engineering, Bilkent University, Ankara, Turkey

Convection-reaction equation based MREPT (cr-MREPT) is known to reconstruct electrical properties (EPs) also in transition regions where EPs vary. However it results in artifacts in regions where “the convective field” is low. Use of multichannel receive data with different and possibly non-overlapping low convective field regions provides the opportunity for artifact free reconstruction of EPs. In this study, a method is developed for combining multichannel receive data to image EPs in a “local region of interest”. The method has been successfully applied in both phantom and healthy human brain experiments.

Jaewook Shin¹, Min-oh Kim¹, Narae Choi¹, and Dong-Hyun Kim^1
1Electrical and Electronic Engineering, Yonsei University, Seodaemun-gu, Seoul, Korea

Magnetic resonance electrical properties tomography (MREPT) is currently being investigated for many clinical applications. However, MREPT suffers from statistical noise and boundary artifact. Especially, the noise amplification in MREPT is occurred due to the calculation of the Laplacian operator. To overcome this EPT error, filtering or fitting based technique was introduced. In this study, low pass filter (LPF) based EPT reconstruction method without the Laplacian operator is proposed.

Jiaen Liu¹, Yicun Wang¹, Xiaotong Zhang¹, Pierre-Francois Van de Moortele², 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) is recently introduced for imaging the electrical properties of tissue using MRI. The EPT equation can be transformed into a partial different equation (PDE) and solved using numerical PDE solutions. The advantage of the PDE-based approach is improved EP result near the boundary and enhanced robustness against noise contamination. A multi-channel transceiver RF coil was utilized to provide multiple transmit B1 field for solving the PDE. The method does not require assumption of equal transmit and receive RF phase, and eliminates the need of pre-assigned boundary condition by using multiple excitations, beneficial for in vivo applications.

Seung-Kyun Lee^1
1GE Global Research, Niskayuna, NY, United States

Noise amplification by Laplacian operation on a noisy input RF map is an important limiting factor in the SNR of MR-based electrical properties tomography (MREPT). We show that among all linear Laplacian kernels, the one based on the Savitzky-Golay second-order derivative kernel has the least amount of noise amplification. A method to construct such a kernel for an arbitrary three-dimensional ROI is presented, and its performance is compared with other Laplacian kernels used in literature.

YuPeng Liao¹, Sandro Romanzetti¹, Vincent Gras¹, DengFeng Huang¹, and N. Jon Shah^1,2
1Institute of Neuroscience and Medicine-4, Forschungszentrum Juelich, Juelich, Germany, ²JARA-Faculty of Medicine, RWTH Aachen University, Aachen, Germany

MRI can be used to non-invasively estimate in vivo electrical conductivity, tissue water content, tissue sodium content (TSC) . In this study we compare proton density (PD), TSC and conductivity distributions in healthy volunteers. This combined study of the three maps, conductivity, PD, and TSC for the first time, allows us to describe the contribution of the tissue water contents and tissue ion concentration (Na+) to the tissue conductivity data and empirical relationship models may be used to predict dielectric properties in other frequency and tissues.

Yicun Wang¹, Xiaotong Zhang¹, Jiaen Liu¹, Pierre-Francois Van de Moortele², and Bin He^1,3
1Department of Biomedical 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) is able to provide quantitative maps of electrical properties (EP) using MRI, which holds promise in early cancer diagnosis. Currently, the Helmholtz Equation-based EPT approaches ignore gradient information on the tissue boundary, causing severe EP reconstruction artifacts; other studies require either a priori boundary conditions that are subject to error, or dielectrical padding that may cause difficulties in practical settings. In this study, we proposed a total variance promoting gradient-based algorithm that does not bear the aforementioned limitations. Simulation and phantom experiment at 7T have demonstrated feasibility of this new approach.

Ji Eun Kim¹, Saurav ZK Sajib¹, Woo Chul Jeong¹, Hyung Joong Kim¹, Oh In Kwon², and Eung Je Woo^1
1Kyung Hee University, Yongin, Gyeonggi, Korea, ²Konkuk University, Seoul, Korea

Electromagnetic field distribution of biological system can be imaged from magnetic flux density which was induced by externally injected current through the electrodes. Electromagnetic field is affected by injected current and electrical conductivity of tissues, electrode type and position are important factors for determining voltage and current density. Signal intensity of current density is proportional to magnetic flux density which can be measured by MREIT. MREIT has potential to provide high-resolution electromagnetic field distribution inside the liver. Using a three-dimensional FEM model, we estimate current pathway and electrical field distribution to detect liver abnormalities at three different current injection methods.

Saurav ZK Sajib¹, Woo Chul Jeong¹, Ji Eun Kim¹, Hyung Joong Kim¹, Oh In Kwon², and Eung Je Woo^1
1Kyung Hee University, Yongin, Gyeonggi, Korea, ²Konkuk University, Seoul, Korea

Estimation of brain anisotropic conductivity has a potential in analysis of interactions between electromagnetic fields and biological systems, such as prediction of current pathway in tDCS and DBS. In this study, we reconstruct anisotropic conductivity distribution of canine brain by combing information from the DTI and MREIT imaging. Since the MREIT technique is used for visualizing high-resolution isotropic conductivity at low frequency. Combing the measured water self-diffusion from DTI and z-component of magnetic flux density from MREIT, it is possible to obtain effective-conductivity-to-diffusivity ratio (ECDR) in every voxels which provides precise information for brain anisotropy comparing to global scaling factor.

Ulrich Katscher¹, Hiroyuki Abe², Marko K Ivancevic³, and Jochen Keupp^1
1Philips Research Europe, Hamburg, Germany, ²Medical Center, University of Chicago, Chicago, Illinois, United States, ³Philips Healthcare, Best, Netherlands

According to ex vivo studies, breast tumors exhibit a significantly altered conductivity, which opens the chance to increase MRI specificity of breast tumor characterization. Conductivity can be measured in vivo using “Electric Properties Tomography” (EPT). With EPT, a trend towards a correlation between conductivity and malignancy was indicated in initial breast tumor studies, however, without statistical significance. This study tests a statistically significant correlation between conductivity and malignancy. Lesion volume derived from pre/post contrast subtraction images was used as a priori information to stabilize EPT reconstruction. According to the results obtained, EPT is able for tumor malignancy staging.

Min Jung Kim¹, Soo-Yeon Kim¹, Dong-Hyun Kim², Jaewook Shin², and Eun-Kyung Kim^1
1Yonsei University, Seoul, Korea, ²Yonsei University, Korea

There is an inverse correlation between conductivity and ADC in invasive breast cancer. Necrosis may have an effect to increase conductivity since cell membranes that are barriers for ionic movement are disrupted by necrosis.

Quantitative Susceptibility Mapping

Monday 1 June 2015

Simon Daniel Robinson¹, Wolfgang Bogner¹, Barbara Dymerska¹, Pedro Cardoso¹, Günther Grabner¹, Xeni Deligianni², Oliver Bieri², and Siegfried Trattnig^1
1High Field MR Centre, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Vienna, Austria, ²Division of Radiological Physics, Department of Radiology, University of Basel Hospital, Basel, Switzerland

We present a phase-sensitive method for the combination of data from array coils for which there is no volume reference coil, such as PTx and some UHF receive arrays. A very fast variable echo-time (vTE) reference scan with a sub-millisecond TE is used to measure the complex sensitivity of the array. Dividing high resolution, T2*-weighted (e.g. SWI) data by the vTE scan matches the phase of the array elements, allowing the complex signals to be combined. This approach is fast, robust, requires no phase unwrapping of phase and outperforms the rival methods tested.

Saifeng Liu¹, Yongquan Ye², Sagar Buch³, and E. Mark Haacke^1,2
1The MRI Institute for Biomedical Research, Waterloo, Ontario, Canada, ²Department of Radiology, Wayne State University, Detroit, Michigan, United States,³School of Biomedical Engineering, McMaster University, Hamilton, Ontario, Canada

When GRAPPA is used for parallel imaging, magnitude and phase images will be reconstructed for each channel of the array coil before being combined for the final images. Simple weighted averaging of the complex data may lead to a particular singularity artifact in phase images, attributed to the variation in the baseline phase components (φ0) between different channels. In this study, we propose an algorithm in which the coil sensitivity induced phase component is modelled as 3D linear function, and can be corrected effectively in k-space domain.

Joseph Dagher^1
1Department of Medical Imaging, University of Arizona, Tucson, AZ, United States

Multi-Echo Gradient Echo phase measurements are important in various MR applications. Bipolar acquisitions of the MEGE echoes in a given TR allow for shorter inter-echo spacing and higher SNR efficiency than monopolar acquisitions, but introduce an unknown spatially varying phase between even and odd echoes. We propose here a strategy for reconstructing the underlying phase, as acquired with an array of receive coils using bipolar echoes, without requiring a reference scan. Our method separates the phase due to the underlying physiology from the phase offsets of the receive coils and the even-odd echoes, using a unifying Maximum-Likelihood framework.

Diana Khabipova¹ and José P. Marques^1
1CIBM, Lausanne, Vaud, Switzerland

Quantitative susceptibility mapping (QSM) is sensitive to iron and myelin but pre-processing need to be performed in order to retrieve the QSM maps. The background removal step was analyzed and the performance and quality of some of the state of art background removal techniques in the estimated susceptibility of the brain cortex was compared in this study for the first time. While all methods have the expected high susceptibility values of primary sensory areas in the layer close to the white matter, the main differences lay in the outer layer of the brain cortex.

Yoonho Nam¹, Dong-Hyun Kim², and Jongho Lee^1
1Department of Electrical and Computer Engineering, Seoul National University, Seoul, Seoul, Korea, ²Department of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea

Recent studies have demonstrated that signals from three water compartments in white matter have different B0 orientation dependent frequency offsets. This observation improved data fitting results in GRE-based MWI by using complex signal fitting as compared to magnitude signal fitting. However, the complex signal fitting approaches applied in the previous studies required several pre-processing steps including a nonlocal background field removal step. Therefore, the results were strongly influenced by them. In this study, we propose a new fitting method that does not require a prior background field removal step. This method shows improvement in parameter estimation.

Wei Li^1,2, Bing Wu³, and Chunlei Liu^4,5
1Research Imaging Institute, University of Texas Health Science Center at San Antonio, San Antonio, TX, United States, ²Ophthalmology, University of Texas Health Science Center at San Antonio, San Antonio, TX, United States, ³GE Healthcare, Beijing, China, ⁴Brain Imaging and Analysis Center, Duke University, Durham, NC, United States, ⁵Radiology, Duke University, Durham, NC, United States

Quantitative susceptibility mapping requires reliable 3D phase unwrapping and background phase removal. Previously, we have developed a method for integrated 3D phase unwrapping and harmonic background phase removal using Laplacian, namely HARPERELLA. In this study, we introduced an improved version of this HARPERELLA method, which allows more intuitive implementation and more robust suppression of the low frequency background phase. Importantly, it also provides the selection between smooth phase throughout the field of view or intact local phase variations around the veins by choosing between Laplacian-based and path-based phase unwrapping.

Debra E. Horng^1,2, Samir D. Sharma¹, Diego Hernando¹, and Scott B. Reeder^1,2
1Radiology, University of Wisconsin-Madison, Madison, WI, United States, ²Medical Physics, University of Wisconsin-Madison, Madison, WI, United States

Background field removal is a necessary step for accurate QSM (susceptibility estimation). We assess the performance of currently used background field removal algorithms (PDF and SHARP), in the context of abdominal imaging. Compared to brain iron, liver iron is spatially diffuse, highly concentrated, and situated closer to air. Large and small containers (6cm and 1.6 diameters) containing gadolinium dilutions, positioned both close and far from air, produced field maps; the dilutions corresponded to susceptibilities of -7.3, -5.6, -3.9, and -2.1 ppm. Both PDF and SHARP result in less than 0.6 Hz/voxel error in the local field estimate.

PINAR SENAY ÖZBAY^1,2, Cristina Rossi¹, Klaas Paul Prüssmann², and Daniel Nanz^1
1Department of Radiology, University Hospital Zürich, Zürich, Switzerland, ²Institute of Biomedical Engineering, ETH Zürich, Zürich, Switzerland

The calculation of quantitative susceptibility maps, QSM, requires removal of non-local fields, e.g., via a convolution of a phase map with a Laplacian-kernel, multiplication with an eroded-binary-tissue-mask, and deconvolution. While erosion of the tissue-mask improves QSM image quality, valuable information is lost. The goal of this study was to regionally vary the size of the Laplacian convolution kernel based on the magnitude of the Frequency-Offset-Gradient-(FOG), which tends to be higher in regions close to the cavities, where air-tissue-susceptibility-artifacts are severe. Susceptibility-maps calculated with the variable-kernel approach revealed less-artifact, and gave more realistic information around the brain-tissue close to the cavities.

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

Instant parameter sweep for L2-regularized Quantitative Susceptibility Mapping, combined with efficient/automatic spatial buffer handling reducing wrap-around streaking.

Hongfu Sun¹, Mahesh Kate², Laura C. Gioia², Derek J. Emery³, Kenneth Butcher², and Alan H. Wilman^1
1Biomedical Engineering, University of Alberta, Edmonton, AB, Canada, ²Neurology, U of Alberta, AB, Canada, ³Radiology, U of Alberta, AB, Canada

We propose a masking dipole inversion and superposition QSM reconstruction method to reduce the artifacts associated with intracranial hemorrhage, in cases of standard susceptibility-weighted imaging using a single long echo.

Hongjiang Wei¹, Wei Li², Nian Wang¹, and Chunlei Liu^1,3
1Brain Imaging and Analysis Center, Duke University, Durham, NC, United States, ²University of Texas Health Science Center at San Antonio, TX, United States, ³Depatment of Radoilogy, School of Medicine, Duke University, Durham, NC, United States

Quantifying susceptibility from the phase image is hampered by the ill-posed dipole filter inversion problem. The imperfect inversion on and near the conical surface results in streaking artifacts in the computed susceptibility maps. In this study, we introduced a novel image-space weighting function to suppress errors induced by imperfect phase measurement and unwrapping. This weighting function is applied in a joint L1 and L2 norm minimization procedure which can be solved rapidly using SPARSA solver. A significantly lower level of streaking artifacts is observed in the resulting susceptibility maps. The results are comparable to those obtained from COSMOS method, and the computation time for the reconstruction is less than one minute for a matrix size of 320×320×204

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

We propose a novel QSM reconstruction method that reduces the artifacts and generates a high quantitative susceptibility map without the regularization term. The method consists of three steps: (I) iterative least square minimization, (II) adaptive edge-preserving filtering to the susceptibility map in the minimization process, and (III) weighted addition of the susceptibility map in k-space before and after filtering. The results of a 3D numerical phantom simulation and healthy volunteer experiment showed that the method may be useful for reducing streaking artifacts and generating a susceptibility map of which quantitative is maintained.

Yan Wen^1,2, Yi Wang^2,3, and Tian Liu^1
1MedImageMetric LLC, New York, New York, United States, ²Biomedical Engineering, Cornell University, Ithaca, New York, United States, ³Radiology, Weill Cornell Medical College, New York, New York, United States

K-space QSM algorithms are computationally inexpensive and easy to implement. But their results usually contains streaking artifacts. Here, we introduce a method that can be applied to an existing k-space results to improve its accuracy and suppress streaking artifacts by constraining the energy of the data in the cone region to the energy of the data in non-cone region, and enforcing structure consistency with sophisticated prior data. This post-QSM method was tested on a gadolinium phantom and an in vivo human brain, and it demonstrated the suppression of streaking artifacts as well as the recovery of cone region data.

Jakob Meineke¹, Julien Senegas¹, Ulrich Katscher¹, and Fabian Wenzel^1
1Philips Research Europe, Hamburg, Hamburg, Germany

Quantitative Susceptibility Mapping is an ill-posed problem that requires two computational steps: background-field removal and dipole inversion. This work focuses on QSM in the brain and describes susceptibility reconstruction using reliable, a priori knowledge from automated segmentation of anatomical ROIs. Background field removal is integrated with dipole inversion in a single-step minimization problem by computing the sources of the magnetic field perturbation which is generated by the tissue susceptibility. This method, dubbed segmentation-enabled dipole inversion (SEDI), results in a more accurate susceptibility reconstruction.

Lijun Bao^1,2, Zhong Chen¹, Peter C.M. van Zijl², and Xu Li^2
1Department of Electronic Science, Xiamen University, Xiamen, Fujian, China, ²Department of Radiology,School of medicine, Johns Hopkins University, Baltimore, MD, United States

Quantitative susceptibility mapping enables non-invasive mapping and quantitative analysis of tissue susceptibility. Interest in this methodology is increasing, because it has the potential to facilitate diagnosis of cerebrovascular and nervous system diseases. However, the reconstruction of magnetic susceptibility constants from local phase information is an ill-posed inverse problem. Therefore, the development of methods to reconstruct accurate susceptibility distributions is important. We present a structural feature based collaborative reconstruction method for quantitative susceptibility mapping. Experimental results on human brain show that this method can provide high quality images of quantitative susceptibility and improve the reconstruction accuracy.

Yilin Yang¹, Tian Liu², Jianwu Dong³, Pascal Spincemaille⁴, and Yi Wang^4,5
1Department of Electronic Engineering, Tsinghua University, Beijing, Beijing, China, ²MedImageMetric, LLC, New York, NY, United States, ³Department of Automation, Tsinghua University, Beijing, Beijing, China, ⁴Department of Radiology, Weill Medical College of Cornell University, New York, NY, United States,⁵Department of Biomedical Engineering, Cornell University, Ithaca, NY, United States

Dipole inversion is the final step of the QSM algorithm. In this step, the zero cone surface in the dipole kernel makes the field-to-susceptibility inverse problem ill-posed. Current solutions are mostly based on the Bayesian approach. Previous techniques have used weighted L1-norm with binary weights derived from the gradient echo magnitude image or phase image. Taking the information from the distribution of the susceptibility gradient into account could improve the reconstructed image. And L2-norm converges faster than L1-norm. Therefore, we employ reweighted L2-norm to specify the distribution to Gaussian. The results of this novel Distribution Specified Dipole Inversion (DSDI) method demonstrate an enhancement of QSM reconstruction and a significant shortening in calculation time.

Zhiwei Zheng¹, Shuhui Cai¹, Congbo Cai², and Zhong Chen^1
1Department of Electronic Science, Xiamen University, Xiamen, Fujian, China, ²Department of Communication Engineering, Xiamen University, Xiamen, Fujian, China

Quantitative susceptibility mapping has found many promising research and clinical applications recently. However, efficient reconstruction method of susceptibility map remains necessary. We propose to reconstruct susceptibility map using a piecewise gradient from estimated susceptibility itself. The gradually varied weighting is different from the traditional binary weighting. Our method is compared to morphology enabled dipole inversion (MEDI) method, and the results indicate that our method has a better performance in susceptibility reconstruction.

Olaf Dietrich¹, Seyed-Ahmad Ahmadi², Johannes Levin², Juliana Maiostre², Annika Plate², Armin Giese³, Kai Bötzel², Maximilian F Reiser^1,4, and Birgit Ertl-Wagner^4
1Josef Lissner Laboratory for Biomedical Imaging, Institute for Clinical Radiology, LMU Ludwig Maximilian University of Munich, Munich, Germany, ²Department of Neurology, LMU Ludwig Maximilian University of Munich, Munich, Germany, ³Center for Neuropathology and Prion Research, LMU Ludwig Maximilian University of Munich, Munich, Germany, ⁴Institute for Clinical Radiology, LMU Ludwig Maximilian University of Munich, Munich, Germany

The purpose of this study was to analyze the influence of the regularization parameters on the results of quantitative susceptibility mapping using the superfast dipole inversion (SDI) technique. SDI is performed in the Fourier domain by multiplication with an SDI kernel, which depends on two regularization parameters for the regularization (1) of the Laplace operator inversion and (2) of the unit dipole inversion (in the Fourier domain). Both parameters were varied and the susceptibility of the substantia nigra and the red nucleus were assessed. The results showed a substantial variation of the susceptibility depending on the regularization.

Diego Hernando¹, Debra E. Horng^1,2, Samir D. Sharma¹, and Scott B. Reeder^1,2
1Radiology, University of Wisconsin-Madison, Madison, WI, United States, ²Medical Physics, University of Wisconsin-Madison, Madison, WI, United States

QSM has the potential to probe MR-invisible objects, based on the effect of their magnetic susceptibility on the nearby B0 field in MR-visible regions. However, the fundamental ability to map magnetic susceptibility of MR-invisible objects based on nearby B0 measurements has not yet been characterized. In this work, we address this question with phantom experiments and singular value decomposition analysis. Our results demonstrate that susceptibility measurement may be feasible in MR-invisible regions if the susceptibility distribution can be assumed to be homogeneous over the MR-invisible region of interest. However, susceptibility mapping of spatially-varying susceptibility distributions in MR-invisible regions is an inherently challenging problem.

Sarah Eskreis-Winkler¹, Dong Zhou², Tian Liu³, Ajay Gupta², Susan Gauthier², Yi Wang², and Pascal Spincemaille^2
1Weill Cornell Medical College, New York, NY, United States, ²Weill Cornell Medical College, New York, United States, ³MedImageMetric, LLC, New York, United States

Zero padding is a well-studied interpolation technique that improves image visualization without increasing image resolution. Here, however, we demonstrate the unique effect of performing zero padding in conjunction with the nonlinear Quantitative Susceptibility Mapping (QSM) algorithm. We assess the effects of this combination by evaluating apparent spatial resolution, relative error and depiction of multiple sclerosis (MS) lesions on QSM.

Sina Straub¹, Andreas Wetscherek¹, Mark E. Ladd¹, and Frederik B. Laun^1
1Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany

The effect of p-space imaging is simulated for a voxel filled with myelinated axons compared to a voxel with homogeneous magnetization. P-space can be obtained by sampling k-space with a standard gradient echo sequence and applying additional gradients before the imaging gradients. In the simulations, little to no difference is observed between the MR signal of the voxel with substructure and the one with homogeneous magnetization. P-space MR experiments, however, showed partial volume effects and signal enhancement in phase and magnitude images that are indicative of subvoxel structure.

Ukash Nakarmi¹, Shruti Prasad², Leslie Ying^1,3, Paul Polak², Robert Zivadinov^2,4, and Ferdinand Schweser^2,4
1Dept. of Electrical Engineering, State University of New York at Buffalo, Buffalo, NY, United States, ²Buffalo Neuroimaging Analysis Center, Dept of Neurology, School of Medicine and Biomedical Sciences, State University of New York at Buffalo, NY, United States, ³Dept. of Biomedical Engineering, State University of New York at Buffalo, NY, United States, ⁴MRI Molecular and Translational Imaging Center, Buffalo CTRC, State University of New York at Buffalo, Buffalo, NY, United States

Application of conventional Compressive Sensing (CS) paradigm on sparse signal reconstruction for Quantitative Susceptibility Mapping is investigated. Frontiers and limitations of CS reconstruction in undersampled GRE data with long echo times and strong phase, and its corresponding effects on tissue characteristics and microstructures on reconstructed images for QSM is investigated.

Patrick McDaniel¹, Audrey Fan², Berkin Bilgic³, Jeffrey N. Stout⁴, and Elfar Adalsteinsson^1,4
1Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States, ²Radiology, Richard M. Lucas Center for Imaging, Stanford University, Stanford, CA, United States, ³A. A. Martinos Center for Imaging, Department of Radiology, Massachusetts General Hopsital, Charlestown, MA, United States, ⁴Health Sciences and Technology, Harvard-MIT, Cambridge, MA, United States

Regional quantitative Oxygen Extraction Fraction (OEF) values measured using MRI phase data are clinically valuable, but can only be reliably obtained from blood vessels larger than the acquisition voxel size. Blood vessels beyond this limit produce unreliable measurements because of partial-volume effects. We demonstrate a method for obtaining more reliable OEF measurements beyond this limit by amending the conventional phase-based reconstruction with additional information contained in the magnitude of the complex-valued MR signal. The method is validated on numerical and in vivo data over a range of voxel sizes from ½ to 3 times the vessel diameter.

Paul Polak¹, Robert Zivadinov^1,2, and Ferdinand Schweser^1,2
1Department of Neurology, Buffalo Neuroimaging Analysis Center, State University of New York at Buffalo, Buffalo, NY, United States, ²Molecular and Translational Imaging Center, MRI Center, Clincal and Translational Research Center, Buffalo, NY, United States

Differentiation of hemorrhagic and calcified brain lesions is an important clinical neuroimaging task. The gold-standard technique uses computed tomography, although the use of susceptibility weighted phase images has been the subject of intense research. On the basis that hemorrhages are paramagnetic, and calcifications diamagnetic, the phase of the lesion can be used as the criterion for differentiation. In numerical simulations of an elliptical lesion we demonstrated that the internal phase also depends on the lesion’s orientation in the magnetic field, and thus phase is an unreliable standard for lesion discrimination. Quantitative susceptibility maps derived from the phase images correctly recovered the underlying susceptibility regardless of positioning, and thus are a better differentiator of lesion subtypes.