Karl Landheer¹, Martin Gajdošík¹, and Christoph Juchem^1,2
1Biomedical Engineering, Columbia University, New York City, NY, United States, ²Radiology, Columbia University, New York City, NY, United States

 

Realistic synthetic PRESS spectra were generated for three different echo times for each of the three major vendors. These spectra were then fit to the matched basis set (i.e., the basis set used to generate it), as well as the mismatched basis sets at the same echo time from other vendors, and the matched basis set but with the hard pulse approximation, to investigate how sensitive resulting quantification is to basis sets. It was found that the concentration for low-concentration metabolites is highly susceptible to small changes in basis sets (e.g., GABA varied by 115 ± 188%).

Andres Saucedo¹, Manoj Kumar Sarma¹, Uzay Emir², James Sayre¹, Paul M Macey³, and Michael Albert Thomas^1
1Radiological Sciences, UCLA Geffen School of Medicine, Los Angeles, CA, United States, ²School of Health Sciences, Purdue University, West Lafayette, IN, United States, ³UCLA School of Nursing, Los Angeles, CA, United States

Five-dimensional correlated spectroscopic imaging using concentric ring trajectories (5D COSI-CONCEPT) has been implemented. Since the maximum slew rates required for the concentric ring trajectories are lower than those for Cartesian echo-planar spectroscopic imaging for a given spectral bandwidth and spatial resolution, the eddy current issues are less challenging. The acquired 5D COSI-CONCEPT data was regridded along k_x-k_y and the accelerated k_z-t₁ data was reconstructed using group sparsity.  Reproducibility of the sequence was evaluated in a brain phantom using 3T Prisma and Skyra MRI scanners. Pilot findings showed excellent coefficients of variance and intra-class correlation coefficients for all the six metabolites.  

 

Helge Jörn Zöllner^1,2, Sofie Tapper^1,2, Steve C. N. Hui^1,2, Peter B. Barker^1,2, Richard A. E. Edden^1,2, and Georg Oeltzschner^1,2
1Russell H. Morgan Department of Radiology and Radiological Science, The Johns Hopkins University School of Medicine, Baltimore, MD, United States, ²F. M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, MD, United States

Recent consensus states that linear-combination modeling (LCM) is the preferred method for modeling GABA-edited spectra. However, there is little peer-reviewed literature investigating how this should be performed. Here, we compare various modeling strategies for GABA-edited spectra, with different approaches to macromolecular signal parametrization, baseline stiffness, and fitting range. Variance of GABA+/tCr estimates decreased when prior knowledge was maximized, i.e., when a full fit range was used with a dedicated co-edited MM basis function. These new modeling strategies have been implemented in the MATLAB-based open-source MRS analysis toolbox Osprey (github.com/schorschinho/osprey).

Karl Landheer¹ and Christoph Juchem^1,2
1Biomedical Engineering, Columbia University, New York City, NY, United States, ²Radiology, Columbia University, New York City, NY, United States

Cramér-Rao Lower Bounds (CRLBs) have become the routine method to approximate standard deviations for magnetic resonance spectroscopy. CRLBs are theoretically a lower bound on the standard deviation. Realistic synthetic 3 Tesla spectra were used to investigate the relationship between estimated CRLBs, true CRLBs and standard deviations. It was demonstrated that although the CRLBs are theoretically truly a lower bound on the standard deviation this approximation is valid only as long as the model properly characterizes the data. In the case when the basis set deviates from the measured data it was shown that the CRLBs deviate substantially from standard deviations.

Karl Landheer¹ and Christoph Juchem^1,2
1Biomedical Engineering, Columbia University, New York City, NY, United States, ²Radiology, Columbia University, New York City, NY, United States

It has recently been recommended that typical preprocessing tools, such as linebroadening, zero-filling and apodization (cutting), generally be avoided prior to signal quantification via consensus. To date, little explanation has been provided against these tools which have become commonplace. Here we demonstrate via realistic Monte Carlo simulations that such preprocessing tools may reduce the precision of the extracted parameters and artificially reduce the Cramér-Rao Lower Bounds and provide a theoretical outline for why they should be avoided.

Lasya P Sreepada^1,2, Sam H Jiang¹, Huijun Vicky Liao¹, Katherine M Breedlove¹, Eduardo Coello¹, and Alexander P Lin^1
1Radiology, Center for Clinical Spectroscopy, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, United States, ²Radiology, Center for Biomedical Image Computing and Analytics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, United States

As medical imaging enters the information era, there is a rapidly increasing need for big data analytics. Robust pooling and harmonization of multi-site data across diverse cohorts is critical. We compare the performances of current basic methods with ComBat, an Empirical Bayes method that removes batch effects, in harmonizing 1H Brain MR Spectroscopy (MRS) of healthy controls from 4 sites. Basic harmonization did not bring metabolite means closer together and increased variance, while ComBat successfully removed significant site effects as determined by ANOVA and Levene's tests. These results may improve reproducibility and generalizability of MRS studies, especially in clinical space.

Sana Vaziri¹, Adam Autry¹, Yaewon Kim¹, Hsin-Yu Chen¹, Jeremy W Gordon¹, Marisa LaFontaine¹, Jasmine Graham¹, Janine Lupo¹, Jennifer Clarke², Javier Villanueva-Meyer¹, Nancy Ann Oberheim Bush³, Duan Xu¹, Susan M Chang², Peder EZ Larson¹, Daniel B Vigneron^1,4, and Yan Li^1
1Department of Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, CA, United States, ²Department of Neurological Surgery, University of California, San Francisco, San Francisco, CA, United States, ³Department of Neurology, University of California, San Francisco, San Francisco, CA, United States, ⁴Department of Bioengineering and Therapeutic Science, University of California, San Francisco, San Francisco, CA, United States

Real-time monitoring of enzymatic conversion of [1-¹³C]pyruvate to [1-¹³C]lactate and [1-¹³C]bicarbonate in the brain can be performed using dynamic hyperpolarized ¹³C metabolic imaging. However, signal-to-noise ratios of bicarbonate are significantly lower than lactate and pyruvate, making it difficult to assess metabolic flux, particularly in lesions. Denoising techniques employing rank reduction via the multidimensional extension of singular value decomposition have recently been introduced for conventional MR images, diffusion-weighted images and HP-¹³C metabolic imaging. Here, we investigate the use of two higher-order singular value decomposition denoising techniques on dynamic hyperpolarized ¹³C metabolic images acquired from patients with glioma.

Shuai Huang¹, James J. Lah¹, Jason W. Allen¹, and Deqiang Qiu^1
1Emory University, Atlanta, GA, United States

We propose a Bayesian approach with built-in parameter estimation to perform T₂* mapping from undersampled k-space measurements. Compared to conventional regularization-based approaches that require manual parameter tuning, the proposed approach treats the parameter as random variables and jointly recovers them with T₂* map. Additionally, the estimated parameters are adaptive to each dataset, this allows us to achieve better performances than regularization-based approaches where the parameters are fixed after the tuning process. Experiments show that our approach outperforms the state-of-the-art l₁-norm minimization approach, especially in the low-sampling-rate regime.

Qingyong Zhu¹, Yuanyuan Liu², Zhuo-Xu Cui¹, Ziwen Ke¹, and Dong Liang^1,2
1Research Center for Medical AI, SIAT, Chinese Academy of Sciences, Shenzhen, China, ²Paul C. Lauterbur Research Center for Biomedical Imaging, SIAT, Chinese Academy of Sciences, Shenzhen, China

We propose a novel DenOising induCed iTerative recOnstRuction framework (DOCTOR) to realize fast $$$T_{1\rho}$$$ parameter mapping from under-sampled k-space measurements. The proposed formulation constrains simultaneously intensity-based and orientation-based similarity between the reconstructed images and denoised prior images. Two state-of-art 3D denoising technologies are utilized including NLM3D and BM4D. The reconstruction alternates between two steps of denoising and a quadratic programming attacked by non-linear conjugate gradient method. The parameter maps are created from the reconstructed images using conventional fitting with an established relaxometry model. Through experiments in-vivo $$$T_{1\rho}$$$-weighted brain MRI datasets, we can observe superior image-quality of the proposed DOCTOR scheme.

Linda Bianchini¹, João Santinha^2,3, Francesca Botta⁴, Daniela Origgi⁴, Marta Cremonesi⁴, and Alessandro Lascialfari^1
1University of Pavia, Pavia, Italy, ²Champalimaud Center for the Unknown, Lisbon, Portugal, ³Instituto Superior Técnico, Lisbon, Portugal, ⁴European Institute of Oncology IRCCS, Milan, Italy

This study tested the hypothesis that some MRI-based radiomic features could lose their texture descriptive power below a volume threshold, compromising the robustness of prediction models. The ability of features to discriminate different textures as a function of the tumor volume was investigated on T₂-weighted images of a customized pelvic phantom. The images were acquired on three scanners and considered three different textures, from a finer to a coarser one. The texture discriminative ability was shown to depend on tumor volume, with most features losing this property with a VOI smaller than 1 cm³ at 1.5 T.

Sule Sahin^1,2, Shuyu Tang³, Manushka Vaidya², and Peder E.Z. Larson^2
1Graduate Program in Bioengineering, University of California, Berkeley and University of California, San Francisco, Berkeley, CA, United States, ²Radiology, University of California, San Francisco, San Francisco, CA, United States, ³HeartVista, Inc., Los Altos, CA, United States

An alternate to AUC ratio, fitting pyruvate to lactate rate constants (kPL) can be a powerful tool for quantification of hyperpolarized [1-¹³C]pyruvate studies. In this work, a model was developed to fit kPL values to a novel acquisition method where lactate was acquired with a stack-of-spiral bSSFP sequence. The model was utilized to fit kPL on three sets of in vivo data: healthy rat kidneys, mouse prostate tumors and human kidney tumors. It was shown that the fit kPL values matched those fit using an established GRE fitting method for complimentary GRE-acquired data sets.  

Oscar Jalnefjord^1,2, Louise Rosenqvist¹, Mikael Montelius¹, Lukas Lundholm¹, Eva Forssell-Aronsson^1,2, and Maria Ljungberg^1,2
1Department of Radiation Physics, University of Gothenburg, Gothenburg, Sweden, ²Department of Medical Physics and Biomedical Engineering, Sahlgrenska University Hospital, Gothenburg, Sweden

This study provides initial results on the encoding-time dependence of IVIM parameters in tumor obtained from a combination of flow-compensated and non-flow-compensated diffusion-encoded data. This was made possible by constructing a pulse sequence capable of performing flow-compensated diffusion encoding with variable encoding time, which was validated through phantom measurements.

Mohsen Farzi¹, Irvin Teh¹, Darryl McClymont², Hannah Whittington², Craig A. Lygate², and Jürgen E. Schneider^1
1Cardiovascular & Metabolic Medicine, University of Leeds, Leeds, United Kingdom, ²Cardiovascular Medicine, University of Oxford, Oxford, United Kingdom

Diffusion tensor imaging (DTI) is a valuable technique for interrogating tissue microstructure, but the estimated parameters remain an indirect characterisation of the underlying tissue architecture. For direct measurement of biophysical parameters, we propose a two-compartment model to quantify cardiomyocyte radius, volume fraction, and dispersion. The intra- and extra-cellular space were modelled using a cylinder with Bingham distributed axes and an oblate tensor. The model reduced root mean squared error by 5% compared to DTI, with volume fraction = 60%, radius = 5.8𝜇m, and dispersion in the sheetlet plane = 9°. These parameters could serve as biomarkers for characterisation of cardiomyopathies.

Guangyu Dan^1,2, Yuxin Zhang^3,4, Zheng Zhong^1,2, Kaibao Sun¹, Muge Karaman^1,2, Diego Hernando^3,4, and Xiaohong Joe Zhou^1,2,5
1Center for MR Research, University of Illinois at Chicago, Chicago, IL, United States, ²Department of Bioengineering, University of Illinois at Chicago, Chicago, IL, United States, ³Department of Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, ⁴Department of Radiology, School of Medicine and Public Health, University of Wisconsin-Madison, Madison, WI, United States, ⁵Departments of Radiology and Neurosurgery, University of Illinois at Chicago, Chicago, IL, United States

It has been increasingly reported that the diffusion-weighted MRI signal depends on not only the b-value but also the diffusion time. Investigations of diffusion model parameters on diffusion time can provide information on interaction between water molecules and their environment, thus helping reveal tissue microstructures. In this study, we focused on two special cases of a continuous-time random-walk (CTRW) diffusion model; and investigated the diffusion-time dependency of the CTRW parameters in the human brain. Our results showed significant dependency of the CTRW parameters on diffusion times in the range of 100-1000 ms.

Manushka V. Vaidya¹, Donghyun Hong¹, Sule Sahin¹, Georgios Batsios¹, Pavithra Viswanath¹, Sabrina M. Ronen¹, and Peder E.Z. Larson^1
1Department of Radiology, University of California San Francisco, San Francisco, CA, United States

We introduce a kinetic modeling framework to detect glutamate and 2-hydroxyglutarate (2HG) production from hyperpolarized [1-¹³C]alpha-ketoglutarate (C1aKG). Detection of 2HG in vivo is often confounded with [5-¹³C]alpha-ketoglutarate (C5aKG), a natural abundance peak. We employ the model, based on the solution to differential equations of a three-site model, to separately detect 2HG from C5aKG. To test the model, we used cell lysate data where separate signal peaks of 2HG and C5aKG were experimentally measured. Cases with inputs of 2HG alone, 2HG+C5aKG, and C5aKG alone were evaluated to validate the model.

Qianqian Yang¹ and Viktor Vegh^2,3
1Queensland University of Technology, Brisbane, Australia, ²The University of Queensland, Brisbane, Australia, ³Centre for Innovation in Biomedical Imaging Technology, Brisbane, Australia

Diffusion kurtosis imaging (DKI) is an important tool in tissue microstructure studies. The DKI formula for diffusion-weighted MRI signal decay arises through a high order expansion, and the kurtosis has been shown to be sensitive to changes in tissue microstructure. Interestingly, the kurtosis formula describes deviation away from mono-exponential signal decay, like that previously described for anomalous diffusion. Here, we make a link between anomalous sub-diffusion in tissue and diffusion kurtosis. This direct link enables the apparent diffusivity and kurtosis to be computed easily from the sub-diffusion model parameters, leading to superior white-grey matter contrast compared with standard DKI parameters.

Brendan Moloney¹, Xin Li¹, Eric M. Baker¹, and Charles S. Springer^1
1Advanced Imaging Research Center, Oregon Health & Science University, Portland, OR, United States

Activity MRI [aMRI] simulates diffusion-weighted data with three metabolic and cytometric tissue properties: k_io (s^-1), the mean steady-state cellular water efflux rate constant [measuring cellular metabolic activity], r, the cell density (cells/μL), and V (pL), the average cell volume.  We explore the behavior of the aMRI model, and compare its results with pertinent experimental data.  The model is well-behaved, and matches experimental data with parameter values in near absolute agreement with independent literature measurements.
 

Dominika Ciupek¹, Maryam Afzali², Fabian Bogusz¹, Marco Pizzolato^3,4, Derek K. Jones², and Tomasz Pięciak^1,5
1AGH University of Science and Technology, Kraków, Poland, ²Cardiff University Brain Research Imaging Centre (CUBRIC), School of Psychology, Cardiff University, Cardiff, United Kingdom, ³Department of Applied Mathematics and Computer Science, Technical University of Denmark, Kongens Lyngby, Denmark, ⁴Signal Processing Lab (LTS5), École polytechnique fédérale de Lausanne (EPFL), Lausanne, Switzerland, ⁵LPI, ETSI Telecomunicación, Universidad de Valladolid, Valladolid, Spain

We investigate the effect of the inversion time on the spherical mean technique (SMT) and neurite orientation dispersion and density imaging (NODDI) metrics based on the multi-parametric diffusion MR data. Our findings indicate the orientation dispersion index and mean orientation of Watson distribution remains unchanged for a wide range of TI, while volume fractions and diffusivities show significant changes leading to characteristic "up-and-down" behavior concerning the TI used to fit the model.

Jordan A. Chad^1,2, Ofer Pasternak³, and J. Jean Chen^1,2
1Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada, ²Rotman Research Institute, Baycrest Health Sciences, Toronto, ON, Canada, ³Brigham and Women's Hospital, Harvard Medical School, Boston, MA, United States

It has been found that when the single-shell two-compartment free water fit is initialized with a constant trace of the tissue tensor, the final results maintain this constant tissue-trace. It is therefore important to understand the implications of the constant tissue-trace constraint in order to interpret the results of single-shell free water studies. Here we demonstrate that this constraint results in allotting all variation in isotropic diffusivity to the free water compartment while the tissue tensor is unaffected by isotropic variations. It is further shown that the isotropic compartment is more linearly aligned with quadratic variations in diffusivity.

Jinglei Lv¹, Rui Zeng¹, and Fernando Calamante^1
1School of Biomedical Engineering, The University of Sydney, Sydney, Australia

Building a brain template of fiber orientation distribution (FOD) with Diffusion MRI is crucial for population study and disease research on white matter. The population template and “Fixel” based analysis pipeline is increasingly being used for group-wise statistics. The template generated based solely on symmetric diffeomorphic registration of FOD depicts the group-consistent major fiber bundles; however, spatial specificity is far from optimal in regions near cortical gray matter.  In this work, we explore the possibility to leverage the complementary information from T1, T2 and Diffusion MRI and build an unbiased human brain FOD template with multimodal registration method.