Diffusion - Simulation & Validation

Thursday 4 June 2015

Michiel Kleinnijenhuis¹, Jeroen Mollink¹, Paul Kinchesh², Wilfred W Lam¹, Vitaly L Galinsky³, Lawrence R Frank³, Sean C Smart², Saad Jbabdi¹, and Karla L Miller^1
1FMRIB Centre, University of Oxford, Oxford, United Kingdom, ²Department of Oncology, University of Oxford, Oxford, United Kingdom, ³Center for Scientific Computation in Imaging, University of California San Diego, La Jolla, United States

Fibre tracts are generally assumed to be very coherent in their microstructure. Along the fibres, little hinderence to diffusion would then be expected. Diffusion properties along tracts can be investigated at long diffusion times, where longer length scales can be probed. This study assessed the diffusion time dependence of the apparent diffusion coefficient along the fibres of the human corpus callosum (often thought as the most coherent bundle) with post mortem STEAM DWI. The biophysical substrate of this relation was interpreted with aid of Monte Carlo diffusion simulations in a range of axon models, representing bending, fanning and undulating configurations.

Quinten Collier¹, Jelle Veraart^1,2, Ben Jeurissen¹, Arnold J. den dekker^1,3, and Jan Sijbers^1
1iMinds-Vision Lab, University of Antwerp, Antwerp, Antwerp, Belgium, ²Center for Biomedical Imaging, New York University Langone Medical Center, New York, New York, United States, ³Delft Center for System and Control, Delft University of Technology, Delft, Netherlands

Partial volume effects caused by cerebrospinal fluid are an important issue in diffusion MRI. In this work, we study the free water elimination model by analyzing the Cramér-Rao lower bound (CRLB) of its parameters. We show that through optimizing the acquisition protocol by minimizing the trace of the CRLB, a significant gain in the precision of the parameter estimation can be achieved. Moreover, further analysis indicates that regularization and/or constraints are necessary for parameter estimation from voxels with large CSF fractions and/or low FA values. These theoretical findings are confirmed by both simulation and real data experiments.

Xiaopeng Zhou¹, Ken Sakaie¹, Robert Fox¹, and Mark Lowe^1
1The Cleveland Clinic, Cleveland, OH, United States

Eddy current artifact is different for each scanner model due to different sequence and calibration algorithm in use. It is beneficial to monitor eddy current artifact quantitatively and effectively correct it in a multi-center DTI trial. A quantitative evaluation of eddy current method as part of QA protocol was developed and applied to 27 scanners. It is an effective method to detect scanners with poor eddy current calibration and can monitor the eddy current distortion effects at sites longitudinally. This method also can quantitatively evaluate how effectively an eddy current correction method can reduce this artifact.

Michal Komlosh^1,2, Dan Benjamini^3,4, Alan S Barnett³, Ferenc Horkay³, and Peter J Basser^3
1NICHD/NIH, Bethesda, MD, United States, ²CNRM/USUHS, Bethesda, MD, United States, ³NICHD/NIH, MD, United States, ⁴The Iby and Aladar Fleischman Faculty of Engineering, Tel-Aviv University, Israel

A novel silicon oil-filled Glass Capillary Array phantom and analytical a theoretical pipeline is proposed as a gold standard for calibrating and validating high q-value diffusion MRI experiments. This use of this phantom is demonstrated to calibrate high angular resolution diffusion imaging (HARDI) and double pulsed field gradient (dPFG) MRI experiments.

Pim Pullens¹, Piet Bladt¹, and Paul M Parizel^1
1Radiology, University Hospital Antwerp & University of Antwerp, Antwerp, Antwerp, Belgium

There is a need for highly standardized isotropic diffusion phantoms. Most isotropic phantoms are difficult to manufacture in large quantities, are difficult to obtain, are expensive, contain toxic or flammable substances and/or require careful handling, which makes them unsuitable for use in a clinical setting. For CENTER-TBI, a large European multi-center study to investigate effective treatments for traumatic brain injury in a clinical environment, 35 phantoms were needed. We have created an easy to use, robust, cost-effective, and safe isotropic diffusion phantom, which can be produced in a reproducible way.

Pim Pullens¹, Michael Bach², Bram Stieltjes³, Dirk Smeets⁴, and Paul M Parizel^1
1Radiology, University Hospital Antwerp & University of Antwerp, Antwerp, Antwerp, Belgium, ²Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany, ³Radiology, Universitätsspital Basel, Basel, Switzerland, ⁴icoMetrix, Leuven, Belgium

For CENTER-TBI, a large European multi-center study on Traumatic Brain Injury, diffusion tensor imaging (DTI) is one of the main imaging modalities. 35 clinical sites across Europe will be included to participate in the MRI study. To be able to acquire DTI data that is cross-site comparable it is of fundamental importance to assess the quality and variability of DTI measurements across sites. In this work 35 anisotropic phantoms were produced and evaluated in order to construct a set of baseline measurements. There is a considerable range of FA values between phantoms, but this may be attributed to post-processing and/or partial volume effects.

Xiaopeng Zhou¹, Ken Sakaie¹, Josef Debbins², Robert Fox¹, and Mark Lowe^1
1The Cleveland Clinic, Cleveland, OH, United States, ²Barrow Neurological Institute, Phoenix, AZ, United States

In a multicenter trial, QA metrics are necessary to provide confidence in data quality and indicate when scanner repairs may be necessary. Quantitative assessment may provide the ability to detect subtle degradation in scanner performance, offering an opportunity to repair the scanner proactively. An important result would be to avoid costly repeat scans or lost data in a multicenter trial. We propose a receiver operating characteristic (ROC) analysis to establish quantitative thresholds on SNR. We demonstrate the use of this method on phantom data acquired from 27 sites in the SPRINT-MS, a phase II trial of treatment for progressive multiple sclerosis.

Irvin Teh¹, Mahon L Maguire¹, and Jürgen E Schneider^1
1Division of Cardiovascular Medicine, Radcliffe Department of Medicine, University of Oxford, Oxford, United Kingdom

Diffusion MRI is particularly sensitive to the calibration of the gradient system. We present a simple and efficient method for gradient calibration based on comparing the measured apparent diffusion coefficient (ADC) in the x, y and z directions in an isotropic cyclooctane phantom, with known reference values, at a given measured temperature. Pre-calibration measurements of ADC differed by up to 8.5x10^-5 mm²/s between directions, leading to an elevated fractional anisotropy (FA) value of 0.10. Post-calibration values of ADC and FA were 0.9x10^-5 mm²/s and 0.03 respectively. The calibration also benefits geometric accuracy as demonstrated with high-resolution anatomical imaging.

Chien-Lin Yeh^1,2, Ruoyun Ma^1,2, Brain Dale³, Thomas L. Chenevert⁴, Michael A. Boss⁵, and Chen Lin^2
1School of Health Sciences, Purdue University, West lafayette, Indiana, United States, ²Radiology and Imaging Sciences, Indiana University School of Medicine, Indianapolis, Indiana, United States, ³Siemens Medical Solutions, North Carolina, United States, ⁴Department of Radiology, University of Michigan Health System, Michigan, United States,⁵Electromagnetics Division, National Institute of Standards and Technology, Colorado, United States

It is known that gradient non-linearity can introduce an error in the measured ADC when off isocenter application was performed. We measured ADC values at various position offsets in different types of scanners from a single vendor and investigated the possibility of correcting ADC error using gradient field map. The correction produces declining ADC values at X offset (R/L) and an elevated ADC at Z offset (H/F), which are both closer to ADC value at the isocenter. A significantly lower gradient nonlinearity error also be shown in our study for long bore scanner compared to short bore scanner.

Christoph Leuze¹, Raju Tomer², Qiyuan Tian¹, Emily Ferenczi², Dan Spielman¹, Michael Zeineh¹, Karl Deisseroth^2,3, and Jennifer A McNab^1
1Radiology, Stanford University, Stanford, CA, United States, ²Bioengineering, Stanford University, Stanford, CA, United States, ³Psychiatry and Behavioural Research, Stanford University, Stanford, CA, United States

CLARITY is a tissue clearing technique that uses hydrogel-embedding to maintain the structural integrity of the tissue and spatial organization of proteins, nuclei acids and other small molecules while using a detergent to remove the lipids that render the tissue optically opaque. It is expected that biomolecules with an NH2 group will bind to the hydrogel and therefore not be removed by the clearing process. MRI of cleared tissue samples can serve the dual purpose of evaluating the efficacy of the tissue clearing and as a means to learn how much the cleared components, such as lipids, contribute to various types of MRI contrast. Here we demonstrate MR images with a range of different contrast mechanisms in a cleared human brain tissue sample and a cleared, whole, mouse brain.

Qiyuan Tian¹, Christoph W.U. Leuze², Raju Tomer³, Emily Ferenczi³, Michael Zeineh², Karl Deisseroth^3,4, and Jennifer McNab^2
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Radiology, Stanford University, Stanford, CA, United States, ³Bioengineering, Stanford University, Stanford, CA, United States, ⁴Psychiatry and Behavioral Sciences, Stanford University, Stanford, CA, United States

We present 3D structure tensor analysis on CLARITY data as a potential method for validating diffusion MRI techniques.

Modeling & Microstructure

Thursday 4 June 2015

Van Thu Nguyen¹, Farshid Sepehrband¹, Othman Alomair¹, Suyinn Chong², Karine Mardon¹, Quang Tieng¹, Graham Galloway¹, and Nyoman Kurniawan^1
1Centre for Advanced Imaging, The University of Queensland, Brisbane, QLD, Australia, ²Mater Research Institute, The University of Queensland, Brisbane, QLD, Australia

While some studies have used NODDI (Neurite Orientation Dispersion and Density Imaging) to analyze human diffusion data, there has not been any study on animal models. This study fitted NODDI model to two HARDI (High-angular Resolution Diffusion-weighted Imaging) shell data (2 b-values and 30 gradient directions) obtained from the wild type mouse brains. The parameter maps obtained with NODDI reflect known brain anatomy and are consistent with fractional anisotropy (FA) maps obtained with DTI processing, and can be considered a complement to DTI by offering additional information of neurite density and their orientation dispersion in both gray and white matters.

Farshid Sepehrband^1,2, Kristi A Clark³, Jeremy F. P Ullmann¹, Nyoman D Kurniawan¹, Gayeshika Leanage¹, David C Reutens¹, and Zhengyi Yang^1,4
1Centre for Advanced Imaging, University of Queensland, Brisbane, Queensland, Australia, ²Queensland Brain Institute, University of Queensland, Brisbane, Queensland, Australia, ³Institute for Neuroimaging and Informatics, University of Southern California, Los Angeles, California, United States, ⁴School of Information Technology and Electrical Engineering, University of Queensland, Brisbane, Queensland, Australia

This work describes a quantitative approach to obtain absolute tissue density, focusing on fiber density, directly from diffusion magnetic resonance imaging. Such measures can be used to map brain fiber and cell densities and to quantitatively characterize disease progression. The proposed technique uses prior knowledge of myelin and cell membrane densities to correlate relative intra-cellular and intra-neurite density values obtained from diffusion magnetic resonance imaging to absolute tissue density values. The proposed method is based on the NODDI (neurite orientation distribution and density imaging) technique, which can be scanned with clinically feasible acquisition protocol.

Dan Benjamini^1,2 and Peter J Basser^1
1National Institute of Health, Bethesda, MD, United States, ²Tel Aviv University, Tel Aviv, Israel

We present an analytical framework to measure the nonparametric joint radius-length (R-L) distribution of an ensemble of finite cylindrical pores, and more generally, the eccentricity distribution of anisotropic pores. Employing a 3-D d-PFG acquisition scheme, we first obtain both the marginal radius and length distributions, and then use these to constrain and stabilize the estimate of the joint radius-length distribution, using a feasible number of acquisitions. Axons are known to exhibit local compartment eccentricity variations upon injury; the extent of the variations depends on the severity of the injury. Reconstructing the eccentricity distribution may provide information about changes resulting from injury or development.

Qiang YU¹, Viktor Vegh¹, Kieran O'Brien^1,2, Thorsten Feiweier³, and David Reutens^1
1Centre for Advanced Imaging, University of Queensland, Brisbane, Queensland, Australia, ²Healthcare Sector, Siemens Ltd, Brisbane, Queensland, Australia, ³Siemens Healthcare, Erlangen, Germany

Axon radius and packing density provide information on the role and performance of white-matter pathways. MRI researchers have tried to measure axon radius, however, they assumed the radius of axons follows a gamma distribution or a single axon radius was considered. From diffusion-weighted data, we mapped axon radii and packing density in the corpus callosum by fitting the parameters of the space fractional Bloch-Torrey anomalous diffusion model. We were able to calculate axon radii and packing density without any assumptions of axon radius. We found the values of axon radii to be in good agreement with previous findings.

Yaniv Katz¹, Dan Benjamini^1,2, Peter J Basser², and Uri Nevo^1
1Biomedical Engineering, Tel Aviv University, Ramat Aviv, Tel Aviv, Israel, ²Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), National Institutes of Health, Bethesda, MD, United States

Here we present our successful accurate estimation of pore size distributions, with pores sizes of typical cellular sizes, by the use of clinically applicable pulsed field gradients. We used a double pulsed field gradient (dPFG) diffusion weighted imaging scheme, with analysis that is based on the Multiple Correlation Function framework. We used a well calibrated phantom and demonstrate successful reconstruction of polydisperse distributions, composed of up to 3 different pores radii, with different volume fractions. These results demonstrate that the q value limit (size<1/q) can be broken, and biological cellular sizes can potentially be estimated using a dPFG pulse sequence.

Björn Lampinen¹, Danielle van Westen^2,3, Freddy Ståhlberg^1,2, Jimmy Lätt³, Oskar Hansson⁴, and Markus Nilsson^5
1Dpt. of Medical Radiation Physics, Lund University, Lund, Sweden, ²Dpt. of Diagnostic Radiology, Lund University, Lund, Sweden, ³Imaging and function, Skane University Health Care, Lund, Sweden, ⁴Clinical Memory Research Unit, Clinical Sciences, Malmö, Lund University, Lund, Sweden, ⁵Lund University Bioimaging Center, Lund University, Lund, Sweden

Neurite density imaging (NDI) is a fast method for obtaining the intracellular volume fraction, or neurite density, from neurite orientation dispersion and density imaging (NODDI). By using powder averaging of the signal to induce full orientation dispersion, NDI simplifies the NODDI model and reduces the acquisition and analysis time. We show that NDI produces accurate neurite density maps from data acquired in only five minutes, thus cutting the minimum acquisition time in half. The method is employed here to disambiguate the cause of elevated FA in the hippocampal cingulum in patients with Parkinson’s disease dementia (PDD).

Damien J McHugh^1,2, Penny L Hubbard Cristinacce^1,2, Josephine H Naish^1,2, and Geoff J M Parker^1,2
1Centre for Imaging Sciences, The University of Manchester, Manchester, United Kingdom, ²Biomedical Imaging Institute, The University of Manchester, Manchester, United Kingdom

This work explores the idea that diffusion-weighted signals acquired with different sequence parameters differ in their sensitivity to various changes in microstructural tissue properties. Monte Carlo simulations were used to investigate how sensitivity to changes in cell size, R, volume fraction, f_i, and permeability, , varies with gradient strength,G, and separation, . Sensitivity was found to depend on specific tissue properties and sequence parameters. For the parameters investigated, mean sensitivity to R, f_i and was maximised using G≈80 mT/m and ≈40 ms, G≈30 mT/m and ≈100 ms, and G≈50 mT/m and ≈110 ms, respectively.

Farshid Sepehrband^1,2, Daniel C Alexander³, Nyoman D Kurniawan¹, David C Reutens¹, and Zhengyi Yang^1,4
1Centre for Advanced Imaging, University of Queensland, Brisbane, Queensland, Australia, ²Queensland Brain Institute, University of Queensland, Brisbane, Queensland, Australia, ³Department of Computer Science & Centre for Medical Image Computing, University College London, London, United Kingdom, ⁴School of Information Technology and Electrical Engineering, University of Queensland, Brisbane, Queensland, Australia

The ActiveAx, a model-based technique, fits minimal white matter model to diffusion MRI data to obtain orientationally invariant indices of axon diameter and density. The fitting procedure is a limitation in such parametric approaches, because various independent parameters have a similar effect on the acquired signal, which may affect the precision of the estimated measures. In this work we propose a dictionary learning approach to tackle this hurdle. We tested our method using ex vivo imaging of the mouse brain (with maximum b-value of 105,000 s/mm²), and compared our estimated values with electron microscopy.

Wilfred W Lam¹, Bernard Siow^2,3, Lauren Burcaw⁴, Daniel C Alexander^2,3, Mark F Lythgoe², Karla L Miller¹, and Saad Jbabdi^1
1FMRIB Centre, University of Oxford, Oxford, United Kingdom, ²Centre for Advanced Biomedical Imaging, University College London, London, United Kingdom, ³Centre for Medical Image Computing, University College London, London, United Kingdom, ⁴Department of Radiology, New York University School of Medicine, New York, NY, United States

Diffusion imaging has enormous potential for quantitative measurements of geometric properties that are directly relevant to brain function and pathology. We validate a previously presented diffusion spectrum model of the extra-axonal space using a phantom consisting of randomly packed solid fibers. Measured diffusion spectra and estimated microstructural properties of the phantom are compared with those from Monte Carlo simulations and model fitting. The model accurately captures salient properties of the diffusion spectrum and estimates the microstructural properties of the phantom.

Wilfred W Lam¹, Karla L Miller¹, Michiel Kleinnijenhuis¹, and Saad Jbabdi^1
1FMRIB Centre, University of Oxford, Oxford, United Kingdom

Diffusion imaging has enormous potential for quantitative measurements of geometric properties that are relevant to brain anatomy. We present long diffusion time measurements of white matter in healthy volunteers showing that the apparent diffusion coefficient measured parallel to axons exhibits diffusion time dependence. We fit a nested series of models with varying complexity and characterized their dimensional appropriateness. A model representing axons as impermeable, parallel, prolate ellipsoids was robustly fit to the data. Although axons are known not to be cylindrical, the ellipsoids could represent microscopic wiggling of the axons along their trajectories or fanning around the (mean) longitudinal direction.

Hong-Hsi Lee¹, Lauren M. Burcaw¹, Jelle Veraart¹, Els Fieremans¹, and Dmitry S. Novikov^1
1Center for Biomedical Imaging, NYU Langone Medical Center, New York, New York, United States

Time dependence of the diffusion coefficient, D(t), reflects tissue complexity on a μm scale. Here we show that this information can be recovered even when the duration δ of diffusion gradient pulses is not infinitely narrow, and design the framework to extract these parameters from a realistic clinically measured D(t, δ). Technically, D(t, δ) can be viewed as the low-pass filtered “ideal” D(t). We apply this framework to human brain DTI measurements along white matter tracts, and find that the predicted δ-dependence agrees with experiment without any adjustable parameters, and furthermore obtain the correlation length that matches the distance between varicosities found along axons.

Koji Sakai¹, Jun Tazoe², Hajime Yokota², Thorsten Feiweier³, Kentaro Akazawa⁴, Hiroyasu Ikeno², and Kei Yamada^2
1Kyoto University, Kyoto, Kyoto, Japan, ²Kyoto Prefectural University of Medicine, Kyoto, Japan, ³Siemens AG, Erlangen, Germany, ⁴Johns Hopkins University, Maryland, United States

q-space imaging (QSI) can allow us to measure water molecular displacement in micro-meter order and also there have been several attempts to verify the microstructural changes of normal/abnormal human subjects. Nevertheless, because of this long acqusition time, the clinical applications of QSI might have been largely limitted. Therefore, this study aimed to find a feasible combination of q values to be shortening the QSI acquisition for clinical use. For the evaluation, we employed mean displacement (MD) which derived from q-analysis of water molecular displacement distribution.

Jeremy J Flint^1,2, Brian Hansen³, and Stephen J Blackband^1,4
1Neuroscience, University of Florida, Gainesville, Florida, United States, ²UF McKnight Brain Institute, Gainesville, Florida, United States, ³Center for Functionally Integrative Neuroscience, Aarhus University, Aarhus, Denmark, ⁴National High Magnetic Field Lab, Tallahassee, Florida, United States

Recent studies have reported a technique in which restriction spectrum imaging data is processed such that signal generated by the restricted water pool is isolated from that generated by the hindered pool. This method, purported to offer contrast sensitive to cell numbers, is presented as a useful tool for detecting the neoplastic proliferation of cells which occurs during tumor growth. We apply this technique to microimaging data collected in the CA1 region of the rat hippocampus. Our results support claims made regarding cellularity-based contrast as we report enhancement in the cell-dense stratum pyramidale which is not observed in adjacent laminae.

Andre Bongers¹, Han Shen², Erika Davies¹, and Eric Hau^2,3
1Mark Wainwright Analytical Centre, University of New South Wales, Sydney, NSW, Australia, ²Adult Cancer Program, University of New South Wales, Sydney, NSW, Australia,³Cancer Care Centre, St George Hospital, NSW, Australia

This study investigates the value of Oscillating Gradient Diffusion Weighted Imaging (OGSE DWI) to gain information about tumour radiation therapy response. 6 U87 glioblastoma bearing nude mice were separated into an irradiated and control arm and examined with an in-house developed cos-OGSE (f=200Hz) sequence in a pre-clinical scanner. Resulting ADC maps were statistically investigated and compared to corresponding PGSE ADC maps. Average ADC values from OGSE DWI were significantly higher in tumours than in normal tissue and showed significant increase after radiation therapy. ADC response to therapy in OGSE proved to be significantly stronger and earlier than corresponding PGSE ADCs.

Yuichi Suzuki¹, Kouhei Kamiya¹, Masaki Katsura¹, Harushi Mori¹, Akira Kunimatsu¹, Akitake Mukasa², Katsuya Maruyama³, Yasushi Watanabe¹, Takeo Sarashina¹, Keniji Ino¹, Masami Goto¹, Jiro Sato¹, Keiichi Yano¹, Nobuhito Saito², and Kuni Ohtomo^1
1Department of Radiology, The University of Tokyo Hospital, Bunkyo-ku, Tokyo, Japan, ²Department of Neurosurgery, The University of Tokyo Hospital, Bunkyo-ku, Tokyo, Japan, ³Siemens Japan K.K., Tokyo, Japan

In this preliminary research, we applied NODDI analyses to brain tumors and compared its ability to represent tissue microstructural information with those of DTI and DKI.

Kohei Tsuruta^1,2, Ryusuke Irie², Masaaki Hori², Issei Fukunaga^1,2, Yoshitaka Masutani³, Kuohei Kamiya⁴, Akira Nishikori^1,2, Mariko Yoshida², Michimasa Suzuki², Masakazu Miyajima², Madoka Nakajima², Koji Kamagata², Hajime Arai², Atsushi Nakanishi², Shigeki Aoki², and Atsushi Senoo^1
1Tokyo Metropolitan University, Arakawa-ku, Tokyo, Japan, ²Juntendo University School of Medicine, Bunkyo-ku, Tokyo, Japan, ³Faculty of Information Sciences and Graduate School of Information Sciences, Hiroshima City University, Hiroshima, Japan, ⁴Radiology, The University of Tokyo Hospital, Tokyo, Japan

Neurite Orientation Dispersion and Density Imaging (NODDI) is a recently developed technique to evaluate the restricted diffusion. This study was to evaluate diffusional changes of cortico-spinal tract (CST) in patients with idiopathic normal pressure hydrocephalus (iNPH) by NODDI. NODDI produces maps of intra-cellular volume fraction (ICVF), orientation dispersion index (ODI) and isotropic volume fraction (iso VF). In the iNPH patients, ODI of CST significantly decreased. Decreased ODI suggested that axon was compressed and oriented. NODDI could show the microstructural changes on CSTs in the iNPH patients.

Jeroen Mollink¹, Michiel Kleinnijenhuis¹, Stamatios N Sotiropoulos¹, Olaf Ansorge², Saad Jbabdi¹, and Karla L Miller^1
1Nuffield Department of Clinical Neurosciences, FMRIB centre, University of Oxford, Oxford, Oxfordshire, United Kingdom, ²Nuffield Department of Clinical Neurosciences, Neuropathology, University of Oxford, Oxford, Oxfordshire, United Kingdom

Microstructural analysis was performed in the corpus callosum with 1): diffusion time measurements and estimation of fiber orientation from optical microscopy (polarized light imaging).

Ece Ercan¹, Emily T. Wood^2,3, Andrew Webb¹, Daniel S. Reich², and Itamar Ronen^1
1C. J. Gorter Center for High Field MRI, Department of Radiology, Leiden University Medical Center, Leiden, Netherlands, ²Translational Neuroradiology Unit (NINDS), National Institutes of Health, Bethesda, Maryland, United States, ³Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States

Diffusion weighted spectroscopy (DWS) of brain metabolites allows study cell-specific alterations in tissue microstructure by probing the diffusion of intracellular metabolites. In particular, the diffusion properties of the neuronal/axonal N-acetylaspartate have been shown to be sensitive to intraneuronal/axonal damage in a variety of pathologies, such as stroke and multiple sclerosis. Missing so far are empirical assessments of the reproducibility of DWS measures across time and subjects, as well as a systematic investigation of optimal acquisition parameters for DWS experiments, sorely needed for clinical applications of the method. We investigated the inter- and intra-subject variability of empirical and modeled diffusion properties of tNAA. Subsequently, we used a jackknife-like resampling approach to explore the variance of these properties in a set of partial data subsets reflecting different total scan duration.

Wilfred W Lam¹, Bernard Siow^2,3, Sean Foxley¹, Steven A Chance⁴, Rogier B Mars^1,5, Daniel C Alexander^2,3, Mark F Lythgoe², Karla L Miller¹, and Saad Jbabdi^1
1FMRIB Centre, University of Oxford, Oxford, United Kingdom, ²Centre for Advanced Biomedical Imaging, University College London, London, United Kingdom, ³Centre for Medical Image Computing, University College London, London, United Kingdom, ⁴Division of Clinical Neurology, University of Oxford, Oxford, United Kingdom, ⁵Department of Experimental Psychology, University of Oxford, Oxford, United Kingdom

Diffusion imaging has enormous potential for quantitative measurements of geometric properties that are directly relevant to brain function and pathology. We combine diffusion spectrum models of the intra- and extra-axonal space, where axons are idealized as randomly packed, parallel, impermeable cylinders with a distribution of radii. Model predictions are compared with Monte Carlo simulations and the model is fitted to diffusion spectra measured in brain specimens. The model accurately captures salient properties of the simulated diffusion spectra and estimates microstructural properties of the tissue that are in agreement with those found in literature.

Patricia Ulloa¹, Viktor Wottschel², and Martin A. Koch^1
1Institute of Medical Engineering, University of Lübeck, Lübeck, Germany, ²Queen Square MS Centre, UCL Institute of Neurology, University College London, London, United Kingdom

Using two independent pairs of gradient pulses it is possible to obtain non-invasive tissue structure information by comparing the signal between different diffusion gradient orientations. The parallel-perpendicular signal difference is analyzed in vivo and in vitro as an indicator of irregularly shaped pores contributing to the diffusion-weighted signal. The results suggest that the acquired signal in the corticospinal tract contains significant contributions from extra-axonal space, consistent with unexpectedly large pore size estimates from earlier experiments.

Darya Morozov¹, Leah Bar¹, Nir Sochen¹, and Yoram Cohen^1
1The Raymond and Beverly Sackler Faculty of Exact Science, Tel-Aviv University, Tel-Aviv Yaffo, Tel-Aviv Yaffo, Israel

Axon size is an important parameter which affects conduction velocity in neuronal tissues. Recently angular d-PFG MR experiments were used to obtain microstructural information in different neuronal tissues. Since in many cases the ground truth of the studied samples is not known a priori, other and we have used microcapillaries phantoms of different complexity to challenge the microstructural information that can be obtained by modeling the signal in different NMR experiments. In the present study, we tried to evaluate simultaneously the size and direction of such systems from angular d-PFG NMR experiments.

Bibek Dhital^1,2, Elias Kellner³, Marco Reisert³, and Valerij G. Kiselev^3
1German Cancer Consortium (DKTK), Heidelberg, Baden, Germany, ²Department of Diagnostic Radiology, University Medical Center, Freiburg, Baden, Germany, ³University Medical Center, Freiburg, Baden, Germany

Biophysical modeling of diffusion weighted MR signal is an ill-defined inverse problem that needs simplified assumptions regarding different fitting parameters. Using a simple multi-shell diffusion protocol and an isotropic weighted diffusion sequence, we experimentally address two common assumptions made in different models different models: (i) the presence of isotropically restricted water compartment and (ii) the relation between ADC of extra and intraaxonal water in single fiber bundles. The results show that isotropically restricted compartment have a negligible contribution to the signal. Additionally we find that intraaxonal axial diffusivity is greater than extraaxonal axial diffusivity. The results show that isotropically restricted compartment have a negligible contribution to the signal. Additionally we find that intraaxonal axial diffusivity is greater than extraaxonal axial diffusivity.

Masoumeh Dehghani M.¹, Nicolas Kunz², Bernard Lanz¹, and Rolf Gruetter^1,2
1Laboratory for Functional and Metabolic Imaging, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Vaud, Switzerland, ²Centre d’Imagerie Biomédicale, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Vaud, Switzerland

The aim of this study is to address the diffusion characteristics of Ace in the rat brain in vivo. The remarkable sensitivity and spectral resolution of localized 1H MRS at 14T allowed a precise measurement of the diffusion properties of NAA and Ace at very high diffusion weighting The significantly large diffusion of Ace estimated from monoexponential fitting indicates its smaller molecule size. Different diffusion characteristics of Ace in the bi-exponential model at higher b values indicate different in vivo diffusion barriers and cellular restrictions compared to NAA and suggest different intracellular distribution space for Ace in the rat brain.

Nicolas Kunz¹, Stéphane Sizonenko², Petra Susan Hüppi², Rolf Gruetter^1,3, and Yohan van de Looij^4
1CIBM-AIT, EPFL, Lausanne, Vaud, Switzerland, ²Division of Child Growth and Development, University of Geneva, Geneva, Switzerland, ³Department of Radiology, University of Geneva and Lausanne, Lausanne, Switzerland, ⁴University of Geneva, Division of Child Growth and Development, Geneva, Switzerland

It has been shown that magnetic field strength and diffusion time (t_diff) has an influence on diffusion tensor imaging derived parameters, limiting multicentre comparisons with different MR systems. The aim of this work was to investigate effect of B0 and t_diff on NODDI estimated microstructural parameters to better understand this field dependency. We demonstrate the feasibility of reconstructing NODDI model in the rodent brain in-vivo at ultra-high magnetic field using multi-b-value shells acquisition. These preliminary results suggest that FA changes along the CC are not only due to differences in axonal diameter but also to axonal orientation dispersion differences as depicted by NODDI results.

Diffusion Acquisition

Thursday 4 June 2015

Paul Kinchesh¹, Michiel Kleinnijenhuis², Karla L Miller², and Sean C Smart^1
1Department of Oncology, University of Oxford, Oxford, United Kingdom, ²FMRIB Centre, Nuffield Department of Clinical Neurosciences, University of Oxford, United Kingdom

Diffusion weighted stimulated echo imaging is often used for diffusion encoding with very high b-values and long diffusion times. For long diffusion times a significant bias can be introduced through the diffusion encoding effect of the slice selection gradients. Compensation of this effect has recently been achieved by adjustment of the prescribed diffusion encoding gradients, but the adjustment is specific to the exact choice of experimental parameters. This report demonstrates that the effect can be minimized through reduction of the diffusion encoding effect of the slice selection gradients themselves, thereby maintaining the fidelity of any applied diffusion encoding scheme.

Samo Lasic¹, Henrik Lundell², Casper Kaae Sønderby², Daniel Topgaard³, and Tim B. Dyrby^2
1CR Development, Lund, Skåne, Sweden, ²Danish Research Centre for Magnetic Resonance, Copenhagen University Hospital, Hvidovre, Denmark, ³Physical Chemistry, Lund University, Lund, Skåne, Sweden

In multiple diffusion encoding sequences with stimulated echo, additional diffusion weighting introduced by the slice and crusher gradients (together known as butterfly gradients) may significantly disrupt the experimental design and introduce a bias that cannot be easily mitigated. We discuss the bias caused by butterfly gradients in Filter EXchange Imaging (FEXI) and their possible confounding effect on the measurement of apparent exchange rate (AXR). The effect is exemplified by the FEXI data acquired on a yeast suspension. Simulation results indicate that reducing slice thickness in FEXI leads to increasingly underestimated AXR values.

Grant Kaijuin Yang^1,2, Christoph W.U. Leuze², and Jennifer McNab^2
1Electrical Engineering, Stanford University, Stanford, California, United States, ²Radiology, Stanford University, Stanford, California, United States

A stimulated echo (STE) double wave vector (DWV) diffusion imaging sequence provides an SNR advantage over spin echo sequences for short T2 species and long diffusion times. However, implementation of STE-DWV is complicated by the formation of unwanted coherence pathways. This abstract presents a crusher scheme to eliminate the unwanted pathways and their associated image artifacts.

Yogesh Rathi¹, Samo Lasic², Tim Dyrby³, and Carl-Fredrik Westin^4
1Harvard Medical School, Boston, MA, United States, ²Colloidal Resource, Sweden, ³Danish Research Centre for Magnetic Resonance, Denmark, ⁴Harvard Medical School, MA, United States

We propose a novel diffusion imaging scheme called Differential Diffusion Imaging (DDI), which uses a differential of two or more standard single PGSE sequences to boost the sensitivity to restricted diffusion by isolating high frequency components of the spectrum of restricted diffusion. The DDI scheme might potentially allow resolving smaller axon diameters compared to the standard single PGSE sequences.

Maryam Afzali¹, Cem Yolcu^2,3, and Evren Ozarslan^3
1Department of Electrical Engineering, Sharif University of Technology, Tehran, Iran, ²Department of Physics and Astronomy, Università di Padova, Padova, Italy, ³Department of Physics, Bogazici University, Istanbul, Turkey

We employ a model of diffusion-attenuated MR signal for molecules under the influence of a Hookean force. The model can be envisioned as an approximation to the mathematically more difficult restricted diffusion problems. The observed diffusion anisotropy is attributed to the anisotropy of a spring’s stiffness tensor rather than the diffusion coefficient, which is taken to be a scalar. We demonstrate the estimation of the stiffness tensor with a positive definiteness constraint on in vivo data.

Cornelius Eichner^1,2, Stephen F Cauley¹, Julien Cohen-Adad³, Harald E Möller², Robert Turner², Kawin Setsompop¹, and Lawrence L Wald^1
1Martinos Center for Biomedical Imaging, Boston, MA, United States, ²Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, SX, Germany, ³École Polytechnique, University of Montreal, Montreal, QC, Canada

This project aims to remove the noise floor, induced by a Rician noise distribution of magnitude data, in diffusion-weighted imaging with low SNR. We implemented a rephasing algorithm to extract real valued diffusion images from complex datasets. Phase corrected real valued data and traditional magnitude data were analyzed regarding signal averaging, model fitting and ability to resolve crossing fibers. Our results reveal that rephased real valued data eliminate Rician noise bias and, therefore, enable unbiased averaging and diffusion model fitting. For future diffusion applications, this method will help to acquire diffusion data with higher resolutions and/or stronger diffusion weightings.

Wei Liu¹, Kun Zhou¹, and David A. Porter^2
1Siemens Shenzhen Magnetic Resonance Ltd., Shenzhen, Select, China, ²Fraunhofer MEVIS, Institute for Medical Image Computing, Bremen, Germany

In the case of DWI, a common way to reconstruct the partially sampled dataset is to zero fill the missing k-space samples. However, this decreases the image resolution and results in significant blurring in the phase-encoding direction. In this study, we introduce a dual-shot DW-EPI sequence, which achieves full k-space sampling by combining two partially sampled data sets, acquired with opposite phase-encoding gradient polarities. The resulting images have reduced image blurring compared to the standard zero-filled case. In addition, the method of image combination inherently incorporates a standard technique for correcting geometric distortion using the reverse-gradient approach.

Jürgen Finsterbusch^1,2
1Department of Systems Neuroscience, University Medical Center Hamburg-Eppendorf, Hamburg, Germany, ²Neuroimage Nord, University Medical Centers Hamburg-Kiel-Lübeck, Hamburg-Kiel-Lübeck, Germany

Multi-band acquisitions offer a promising approach to accelerate diffusion-weighted acquisitions but require an appropriate coil geometry. For applications in the human cervical spinal cord such coils are not widely available. Here, it is demonstrated that slice acceleration can be achieved nevertheless by using 2D-selective RF excitations to restrict the excitation to the spinal cord in combination with slice-gradient blips to induce different shifts in the image for the different slices excited. The feasibility of this approach is demonstrated in phantoms and the human cervical spinal cord in vivo.

Xiaodong Ma¹, Zhe Zhang¹, Yishi Wang¹, Erpeng Dai¹, and Hua Guo^1
1Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, China

In order to achieve high resolution spine DTI, the 2D-navigated interleaved EPI combined with a proposed reconstruction method, Shot Encoded Parallel-imaging Technique (SEPARATE), is used in this study. Cervical spine and lumbar spine DTI results show the improved resolution and reduced distortion compared to the single-shot EPI DTI. Furthermore, the proposed method is less sensitive to the mismatch between image data and navigator than the image domain phase correction.

Congyu Liao¹, Hongjian He¹, Song Chen¹, Merry Mani², Mathews Jacob², Vincent Magnotta², and Jianhui Zhong^1
1Center for Brain Imaging Science and Technology, Zhejiang University, Hangzhou, Zhejiang, China, ²University of Iowa, Iowa, United States

In this study, we proposed a motion-compensated Iterative Self-consistent Parallel Imaging (SPIRiT) acceleration scheme with analytical Q-Ball imaging (QBI) based reconstruction4 to obtain orientation distribution function (ODF) with high spatial and angular resolution. This metric demonstrates improved quality with reduced noise and motion artifacts compensated with the method.

Xiaodong Ma¹, Bida Zhang², Zhangxuan Hu¹, Trong-Kha Truong³, Allen W. Song³, and Hua Guo^1
1Department of Biomedical Engineering, Tsinghua University, Beijing, China, ²Healthcare Department, Philips Research China, Shanghai, China, ³Brain Imaging and Analysis Center, Duke University, Durham, North Carolina, United States

An intuitive phase correction approach, SENSE+CG, has been proposed and implemented on navigator-free spiral DWI.In this study, regularization was introduced into the SENSE+CG algorithm to solve the semi-convergence problem. The in vivo results show that it can generate diffusion weighted images with fast and stable convergence, which is helpful to determine the stopping criterion without manual intervention.

Mark H Sundman¹, Hing-Chiu Chang¹, Dan Xu², Arnaud Guidon³, and Nan-kuei Chen^1
1Brain Imaging and Analysis Center, Duke University Medical Center, Durham, North Carolina, United States, ²Global MR Applications and Workflow, GE Healthcare, Waukesha, Wisconsin, United States, ³Global MR Applications and Workflow, GE Healthcare, Boston, Massachusetts, United States

This research employs a navigated multi-shot DWI, termed Navigator-MUSE, to enhance image quality. This novel 16 shot DWI technique is capable of improving spatial resolution (0.8 x 0.8 x 0.8 mm3) while also eliminating aliasing artifacts and geometric distortions. We demonstrate that Navigator-MUSE is capable of minimizing aliasing artifacts due to brain pulsation in highly susceptible regions like the brain stem without significant scan-time penalties.

Arturo Cardenas-Blanco¹, Julio Acosta-Cabronero¹, Martin Kanowski², Joern Kaufmann², Claus Tempelman², Stefan Teipel³, and Peter J Nestor^1
1Brain plasticity and neurodegeneration, German Center for Neurodegenerative Diseases (DZNE), Magdeburg, Germany, ²Department of Neurology, Otto-von-Guericke University, Magdeburg, Germany, ³German Center for Neurodegenerative Diseases (DZNE), Rostock, Germany

DTI has become a standard research tool to probe the structural organisation of neural tissue. Sub-optimal acquisitions or processing methodologies can, however, compromise the reliability of results. The aim of this study was to challenge the assumption that enabling 30 diffusion-encoding orientations would be sufficient for obtaining unbiased DTI measurements with standard methods. Results showed strong effects in DTI TBSS group analyses that were head-/FOV-positioning dependent; suggesting, therefore, that group studies with poor acquisition standardisation could lead to spurious results.

Kurt Schilling¹, Yurui Gao¹, Iwona Stepniewska², Ann S Choe¹, Bennett A Landman³, and Adam W Anderson^1
1VUIIS, Vanderbilt University, Nashville, TN, United States, ²Psychology, Vanderbilt University, Nasvhille, United States, ³Electrical Engineering, Vanderbilt University, Nashville, TN, United States

Here, we characterize the diffusion properties of the squirrel monkey brain. We find the reproducibility of the mean diffusivity, fractional anisotropy, and primary eigenvector is comparable to that of human DTI studies, establishing the validity of quantitative cross-sectional and longitudinal DTI studies on the squirrel monkey. Second, the relationship between in vivo and ex vivo is considered. We confirm that death and fixation causes significant changes to the tissue microstructural properties that have a notable effect on diffusion MRI, specifically a decreased mean diffusivity and an increased fractional anisotropy. Finally, we provide the normal values of diffusion indices in a variety of both white and gray matter regions of interest. This study serves as the basis for using the squirrel monkey for diffusion MRI studies, supporting the use of ex vivo DTI, as well as subsequent histology, as a means of understanding image contrast seen on the in vivo scans.

Diffusion Processing & Analysis

Thursday 4 June 2015

Mark S Graham¹, Ivana Drobnjak¹, and Hui Zhang^1
1Department of Computer Science and Centre for Medical Image Computing, UCL, London, United Kingdom

This work highlights issues with the current practice for correcting eddy-current (EC) distortions on moderately high b-value data and demonstrates their mitigation with a simple alternative. Both techniques are evaluated on real and simulated data, and the importance of EC correction for estimating microstructure is illustrated with the NODDI model. We demonstrate that correcting moderately high b-value data with standard EC correction techniques introduces distortion that compromises the anatomical correspondence between the DWIs and leads to questionable estimates of microstructural features. We show our alterative circumvents these issues and provides good correction.

David Rotenberg¹, Peter Savadjiev², Yogesh Rathi², Aristotle Voineskos^3,4, and M. Mallar Chakravarty^5,6
1Research Imaging Centre, Centre for Addiction and Mental Health, Toronto, Ontario, Canada, ²Laboratory of Mathematics and Imaging, Harvard Medical School, MA, United States, ³Centre for Addiction and Mental Health, Ontario, Canada, ⁴Department of Psychiatry, University of Toronto, Ontario, Canada, ⁵Cerebral Imaging Centre, Douglas Mental Health University Institute, Quebec, Canada, ⁶Department of Psychiatry, McGill University, Quebec, Canada

Diffusion Weighted imaging data are typically collected with echo planar imaging (EPI) sequences, prone to geometric distortion in the phase-encode direction. Geometric distortion in EPI images can be minimized by non-linear registration to an anatomical reference image, or by using b0 field maps. In this work we perform an investigation of both field mapping and non-linear registration distortion correction techniques, to evaluate their effect on standard downstream diffusion tensor imaging metrics and tractography. We also present SSECC (Simultaneous Susceptibility and Eddy Current Correction),which corrects for the unique eddy-current and geometric distortions affecting each diffusion image, in an integrated fashion. Tractography, and Tract-Based Statistics results indicate that despite having received little attention in the literature, geometric distortion may have a profound impact on both tractography and diffusion metrics, specifically FA, which appears to be substantially affected throughout the imaging volume.

Yishi Wang¹, Zhe Zhang¹, Xue Zhang¹, Xuesong Li¹, Sheng Xie², Chun Yuan^1,3, and Hua Guo^1
1Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, China, ²Department of Radiology, China-Japan Friendship Hospital, Beijing, China, ³Department of Radiology, University of Washington, Seattle, Washington, United States

Pulsatile brain motion can induce DTI data corruption and affects DTI metrics. Data rejection methods in both image domain and k-space domain have been proposed. In this study, we use large cohort data to evaluate the data rejection rate in k-space in order to investigate the extent to which a specific region suffers from motion. A map of data rejection rate on a standard brain template was obtained from the mean of 77 subjects. Motion affection was quantified and suggestions on data acquisition scheme were given based on the statistical results.

Mahdi Alizadeh^1,2, Pallav Shah², Devon M Middleton^1,2, Chris J Conklin^2,3, Sona Saksena², Scott H Faro^1,2, MJ Mulcahey⁴, Jürgen Finsterbusch⁵, and Feroze B Mohamed^1,2
1Bioengineering, Temple university, Philadelphia, Pennsylvania, United States, ²Radiology, Temple university, Pennsylvania, United States, ³Electrical Engineering, Temple university, Pennsylvania, United States, ⁴Occupational Therapy, Thomas Jefferson University, Pennsylvania, United States, ⁵Systems Neuroscience, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

In this paper we presented a novel technique to segment and distinct ghost artifacts from cord in diffusion tensor spinal cord images. This techniques is multi-stage method included three main steps namely, segmentation, feature extraction and classification. The extracted features from segmented regions classified using Adaptive Neuro Fuzzy Interface System (ANFIS).

Jelle Veraart¹, Florian Knoll¹, Jan Sijbers², Els Fieremans¹, and Dmitry S. Novikov^1
1Center for Biomedical Imaging, NYU Langone Medical Center, New York, NY, United States, ²iMinds - Vision Lab, University of Antwerp, Antwerp, Belgium

MR images are typically distorted with spurious signal that appear near sharp edges in the images. This Gibbs artifact results from the truncation of the k-space. Although the artifacts has a significant impact on the quantification of diffusion MR indices, it is often ignored or only reduced by smoothing the data at the expense of image blurring. The present work demonstrates that extrapolating the data in k-space beyond the measured part by means of second order total generalization variation minimization allows for a suppression of truncation artifacts without compromising resolution or modeling the image as a piecewise constant function.

Samuel St-Jean¹, Guillaume Gilbert², and Maxime Descoteaux^1
1Sherbrooke Connectivity Imaging Lab (SCIL), Université de Sherbrooke, Sherbrooke, Québec, Canada, ²MR Clinical Science, Philips Healthcare, Markham, Ontario, Canada

Diffusion Weighted Images datasets are acquired at a low spatial resolution due to decreased SNR and increased acquisition time as the voxel size is reduced. Achieving high spatial resolution improves the specificity of reconstructed tracts and diffusion features, which might not be present at a lower spatial resolution. We show that high resolution DWIs are achievable thanks to proper denoising and favorably compare to the HCP dataset, while still being feasible in 13 minutes on a standard 3T clinical scanner. This could reveal new anatomical details, which are not achievable at the spatial resolution currently used in diffusion MRI.

Xi Peng¹, Shanshan Wang¹, Yuanyuan Liu¹, and Dong Liang^1
1Paul C. Lauterbur Research Centre for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Shenzhen, Guangdong, China

Low SNR is a significant problem in diffusion tensor imaging. Recent methods using sparse or low rank models usually denoise in image space. One has to go through an estimation chain (i.e., image¡ú tensor¡ú eigen-value¡úFA) to obtain the FA map, which may cause error propagation. This work proposes to use the model-based method for DTI denoising. Notably, we creatively penalize the non-smoothness of the tensor and the nonlinear FA simultaneously. To enable this, we calculate FA values from elements of the tensors directly without computing the eigen-values. Experiments were conducted and show promising results in heavy noise case.

Alexander M. Cerjanic^1,2, Joseph L. Holtrop^1,2, and Bradley P. Sutton^1,2
1Bioengineering, University of Illinois at Urbana-Champaign, Urbana, Illinois, United States, ²Beckman Institute of Advanced Science and Technology, University of Illinois at Urbana-Champaign, Urbana, Illinois, United States

The Intravoxel Incoherent Motion (IVIM) model allows for the quantification of flow in the microvasculature through the use of diffusion weighted imaging. While IVIM has clinical potential, estimating parameter maps from magnitude images is extremely SNR demanding due to Rician noise. Using a maximum penalized likelihood estimator (MPLE), reliable parameter maps can be obtained at high resolution in reasonable times. Parameter maps obtained from a volunteer processed with both the MPLE method and conventional least squares estimates are shown. Comparisons between IVIM parameters in gray matter and white matter demonstrate the promise of MPLE as compared to least-squares estimation.

Xinyuan Zhang¹, Man Xu¹, Zhe Zhang², Hua Guo², Fan Lam³, Zhipei Liang³, Qianjin Feng¹, Wufan Chen¹, and Yanqiu Feng^1
1Biomedical Engineering, Guangdong Provincial Key Laborary of Medical Image Processing, Southern Medical University, Guangzhou, Guangdong, China, ²Biomedical Engineering, Center for Biomedical Imaging Research,Tsinghua University, Beijing, Beijing, China, ³Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Urbana, Illinois, United States

Diffusion-weighted (DW) magnetic resonance imaging is widely used in clinic and research because of its ability to characterize the diffusion of water molecules within tissue. However, the DW images are usually affected by severe noise especially at high resolution and high b values, and the low signal-to-noise ratio may degrade the reliability of the subsequent quantitative analysis. Recently, a patch-based higher-order singular value decomposition (HOSVD) method was proposed to denoise MR images and demonstrated to outperform the well-known BM4D method. Compared with the conventional T1-, T2- and proton density (PD)-weighted images, DW images may contain more redundant information because that they are usually highly correlated across different diffusion directions. In this work, we proposed to simultaneously exploit the redundant information along diffusion directions and across spatial domain by using HOSVD in denoising DW images.

Anna Auria¹, Eric Canales-Rodriguez^2,3, Yves Wiaux⁴, Tim Dirby⁵, Daniel Alexander⁶, Jean-Philippe Thiran^7,8, and Alessandro Daducci^1,8
1Signal Processing Lab (LTS5), EPFL, Lausanne, Switzerland, ²FIDMAG Germanes Hospitalàries, Barcelona, Spain, ³Centro de Investigacion Biomédica en Red de Salud Mental, CIBERSAM, Spain, ⁴Institute of Sensors, Signals and Systems, Heriot-Watt University, Edinburgh, United Kingdom, ⁵Danish Research Centre for Magnetic Resonance, Copenhagen University Hospital Hvidovre, Denmark, ⁶Department of Computer Science and Centre for Medical Image Computing, University College London, United Kingdom, ⁷Signal Processing Lab (LTS5), EPFL, Switzerland, ⁸University Hospital Center (CHUV) and University of Lausanne (UNIL), Switzerland

Mapping the microstructure properties of the local tissues in the brain is crucial to understand any pathological condition from a biological perspective. Most of the existing techniques to estimate the microstructure of the white matter assume a single axon orientation whereas numerous regions of the brain actually present a fiber-crossing configuration. The purpose of the present study is to extend a recent convex optimization framework to recover microstructure parameters in regions with multiple fibers.

Luis Miguel Lacerda¹, Jonathan I. Sperl², Marion I. Menzel², Gareth Barker¹, and Flavio Dell'Acqua^1
1Department of Neuroimaging, The Institute of Psychiatry, Psychology & Neuroscience, King's College London, London, Denmark Hill, United Kingdom, ²GE Global Research, Munich, BY, Germany

Diffusion Spectrum Imaging(DSI) is a very complex technique that requires extensive validation before it can be well established in clinical practise. In this study, we simulated different tissue configurations, sampling schemes and processing steps to evaluate the performance of DSI. From derived simulations it was possible to identify specific configurations where DSI is unable to characterize diffusion without artifacts, namely aliasing caused by fast diffusion components. Furthermore, processing of orientation distribution functions in these environments may lead to generation of spurious fibres. We reviewed the steps involved in the derivation and analysis of DSI data and explored these limitations.

Divya Varadarajan¹ and Justin P Haldar^1
1Electrical Engineering, University of Southern California, Los Angeles, California, United States

The Funk-Radon and Cosine Transform (FRACT) is a recent linear method for estimating orientation distribution functions (ODFs) from single-shell diffusion MRI data. Compared to previous single-shell ODF estimation techniques, the FRACT offers predictable performance, strong theoretical characterization, and does not require any tissue modeling assumptions (that can confound nonlinear ODF estimation methods when the modeling assumptions are violated). In this work, we propose an extension of FRACT for multi-shell diffusion MRI (MS-FRACT). We show theoretically and empirically that MS-FRACT yields more accurate ODF estimates than conventional FRACT, while still being predictable with strong theoretical characterization, and without requiring tissue-modeling assumptions.

Gabriel Varela-Mattatall¹, Alexandra Tobisch^2,3, Tony Stoecker^2,4, and Pablo Irarrazaval^5,6
1Biomedical Imaging Center, Pontificia Universidad Catolica de Chile, Santiago, Metropolitan District, Chile, ²German Center for Neurodegenerative Diseases, North Rhine-Westphalia, Germany, ³Department of Computer Science, University of Bonn, North Rhine-Westphalia, Germany, ⁴Department of Physics and Astronomy, University of Bonn, North Rhine-Westphalia, Germany, ⁵Biomedical Imaging Center, Pontificia Universidad Catolica de Chile, Metropolitan District, Chile, ⁶Department of Electrical Engineering, Pontificia Universidad Catolica de Chile, Metropolitan District, Chile

Compressed Sensing (CS) has been applied to Diffusion Spectrum Imaging (DSI) in order to accelerate acquistion, unfortunately, is difficult to assure high acceleration factors with the conventional DSI-CS formulation because CS is thought for high resolution problems, which is not the case in dMRI in general. In this work we propose a change over the DSI-CS formulation to improve reconstruction based in applying CS to reconstruct the differences from a tensor fitted to the data. This joint method between tensor fitting and the reconstruction of the differences allows an improvement in the reconstruction.

Marco Reisert¹, Katharina Göbel¹, and Bibek Dhital^1
1Medical Physics, University Medical Center Freiburg, Freiburg, Germany

Diffusion weighted magnetic resonance imaging (DWI) gives a unique opportunity to look inside tissue microstructure of human brain white matter. Usually we try to explain the DWI-signal by identifying the different microscopic components and making assumption about their physical and statistical properties. This leads to analytical models containing the relevant parameters like diffusivities and volume fractions. In this work we want to propose an alternative, more a phenomenological description of the measurement. Instead of stating some analytical model which is statistical derived from some physical assumptions about microstructure statistics, we propose to directly reconstruct the tissue microstructure.

Qiyuan Tian¹, Christoph W.U. Leuze², Ariel Rokem³, and Jennifer A. McNab^2
1Department of Electrical Engineering, Stanford University, Stanford, CA, United States, ²Department of Radiology, Stanford University, CA, United States, ³Psychology, Stanford University, Stanford, CA, United States

We demonstrate coherent patterns of crossing fibers in the in vivo human cerebral cortex using high-angular resolution and high spatial resolution diffusion imaging.

Sudhir K Pathak¹, Catherine Fissell², Deepa Krishnaswamy¹, Sowmya Aggarwal¹, Rebecca Hachey², and Walter Schneider^2
1Bioengineering, University Of Pittsburgh, Pittsburgh, PA, United States, ²Psychology, University Of Pittsburgh, Pittsburgh, PA, United States

There are different diffusion reconstruction algorithms reported in the literature that can be used to assess the micro-structure of white matter tissue in the human brain. Most notably, constrained spherical deconvolution (CSD) is used for single shell (constant b-value) image acquisitions and Generalized Q-sampling imaging (GQI) is used for Diffusion Spectrum Imaging and Multi-shell image acquisitions. A new reconstruction method is proposed that combines CSD and GQI to estimate the spherical harmonics coefficients directly from the diffusion signal. The proposed method produce improved estimates of fiber directions and sharper ODF using CSD and better tractography.

Tom Dela Haije¹, Andrea Fuster¹, and Luc Florack^1
1Mathematics and Computer Science, Eindhoven University of Technology, Eindhoven, Noord-Brabant, Netherlands

In this work we describe and perform the reconstruction of general polynomial diffusion MRI models, with the added constraint that the polynomials are convex. This is done by requiring the Hessian of the model to be sum-of-squares. The resulting optimization problem is shown to be solvable in a reasonable amount of time for the scale of data typical in clinical diffusion MRI acquisitions.

Luis Miguel Lacerda¹, Jonathan I. Sperl², Marion I. Menzel², Gareth Barker¹, and Flavio Dell'Acqua^1
1Department of Neuroimaging, The Institute of Psychiatry, Psychology & Neuroscience, King's College London, London, Denmark Hill, United Kingdom, ²GE Global Research, Munich, BY, Germany

Model-free diffusion imaging techniques enable the reconstruction of the orientation distribution function(ODF) from the diffusion propagator. This function should only recover displacement probabilities consistent with white matter avoiding as much as possible partial volume contaminations. In this study, we propose a new method for ODF computation; it is characterized by restricting integration to a range of displacement probabilities, derived from prior knowledge of the expected physical displacements associated with diffusion along the direction of axons. The proposed method returns better white matter orientation showing higher angular resolution, without the need of min-max normalisation, thus retaining the quantitative nature of ODFs.

Ralf Luetzkendorf¹, Robin M Heidemann², Thorsten Feiweier², Joerg Stadler³, Sebastian Baecke¹, Michael Luchtmann⁴, and Johannes Bernarding^1
1Department for Biometry and Medical Informatics, University of Magdeburg, Magdeburg, Germany, ²Siemens Healthcare, Erlangen, Germany, ³Leibniz Institute for Neurobiology, Magdeburg, Germany, ⁴Department of Neurosurgery, University of Magdeburg, Magdeburg, Germany

To analyze fiber orientation density within gray matter (GM) and at gray-white matter borders (GWMB) spherical de-convolution was applied to 1.4 mm isotropic 7T DWI data leading to high-resolution fiber orientation distribution maps (FOD). Data were post-processed with FSL and MRTrix 0.2. Anatomy, DWI and vector schemes were registered to each other. The high resolution decreased partial volume effects in GM allow for clear differentiation of fiber orientations in GM and GWMB in the ODF maps. Fibers in gray matter are oriented perpendicular to local GM-surfaces. Along the rim of the gyri, sharp bending of fibers can be unambiguously seen at GWMB.

Alberto De Luca^1,2, Marco Castellaro¹, Stefania Montemezzi³, Massimiliano Calabrese⁴, and Alessandra Bertoldo^1
1Department of Information Engineering, University of Padova, Padova, PD, Italy, ²Department of Neuroimaging, Scientific Institute, IRCCS "Eugenio Medea", Bosisio Parini, LC, Italy, ³Radiology Unit, Azienda Ospedaliera di Verona, Verona, Italy, ⁴Neurology Section, Department Of Neurological and Movement Sciences, University Hospital of Verona, Verona, Italy

In this work we present a non-linear multi compartmental model based on spherical deconvolution to fit multi-shell diffusion data. The first two parameters of the model provide parametric maps highly correlated to T1 segmentation (up to 85%), while the last parameter leads to a map of diffusivity useful for lesion detection purposes. The residuals are random dispersed around zero and average coefficients of variation between 4 and 24%. Application of the model to a multiple sclerosis subject show that the diffusivity map is sensible to abnormally diffusing voxels, revealing lesions that are confirmed from a FLAIR scan.

Elena Najdenovska^1,2, Giovanni Battistella^3,4, Constantin Tuleasca^1,5, Philippe Maeder⁴, Alessandro Daducci^2,5, Jean-Philippe Thiran^4,5, Marc Levivier¹, Eleonora Fornari^2,4, and Meritxell Bach Cuadra^2,4
1Department of Clinical Neuroscience, Neurosurgery Service and Gamma Knife Center, Centre Hospitalier Universitaire Vaudois (CHUV), Lausanne, Switzerland, ²Centre d'Imagerie BioMédicale (CIBM), Lausanne, Switzerland, ³Department of Neurology, Mount Sinai School of Medicine, New York, United States, ⁴Department of Radiology, Centre Hospitalier Universitaire Vaudois (CHUV), Lausanne, Switzerland, ⁵Signal Processing Laboratory (LTS5), Ecole Polytechnique Fédérale de Lausanne (EPFL), Switzerland

We propose a novel method for segmenting the thalamus in 7 anatomical group based on local diffusion properties at high angular resolution as given by the Spherical Harmonics representation of the Orientation Distribution Functions. The validation was done for a large dataset including 33 healthy subjects and 2 patients treated for essential tremor with Gamma Knife Surgery. We found a robust clustering pattern for all tested healthy subjects and patients before surgery. Additionally, the surgery target in the patient data, used as gold standard for validation, has proven the emplacement accuracy of the motor-related thalamic group delineated with our method.

Ana Karen Loya-Olivas¹, Mariano Rivera¹, and Ramon Aranda^1
1Computer Science Department, Centro de Investigación en Matemáticas, Guanajuato, Guanajuato, Mexico

We have presented a set of improvements to the algorithm that adapts the dictionary of diffusion functions, modifying and rotating the diffusion tensor profile. Our proposal simplifies the optimization problems using linear approximations in the rotations and size of diffusion profiles are estimated solving simple linear least-squares sub-problems. We evaluated our method performance with benchmark data set and we demonstrated its capabilities using real data.

Pew-Thian Yap¹, Yong Zhang², and Dinggang Shen^1
1Department of Radiology, University of North Carolina, Chapel Hill, North Carolina, United States, ²Department of Psychiatry & Behavioral Sciences, Stanford University, California, United States

An inherent limitation of Spherical deconvolution (SD) in estimating the fiber orientation distribution function (FODF) is that the fiber kernel is assumed to be spatially invariant. This has been shown to result in spurious FODF peaks. This abstract describes a multi-kernel approach for robust estimation of the fiber orientation distribution function. We show that instead of restricting ourselves to one kernel per compartment, it is possible to employ a group of kernels per compartment to cater to possible data variation across voxels. Our results demonstrate that the proposed method significantly improves microstructural and tract estimates.

Fang-Cheng Yeh¹ and Timothy Verstynen^1
1Deparment of Psychology, Carnegie Mellon University, Pittsburgh, Pennsylvania, United States

The organization of white matter pathways defines the essential wiring diagram that acts as a hard constraint on neural processing. To this end we reconstructed fiber orientation distribution functions (fODF) for every white matter voxel in the brain in a stereotaxic space to create an fODF atlas. The fODF atlas offers a representative fiber structure that can be used to conduct optimized fiber tractography through complex fiber crossings to visualize the structural connectivity of the human brain.

Ramon Aranda¹, Mariano Rivera¹, and Alonso Ramirez-Manzanares^1
1Computer Science Department, Centro de Investigación en Matemáticas, Guanajuato, Guanajuato, Mexico

In this work, we present an voxel-wise adaptive diffusion dictionary to estimate on in vivo brain the structure of the axonal fiber populations. Our proposal overcomes the following limitations of the diffusion dictionary-based methods: the limited angular resolution and the fixed shapes for the atom set. The improvements obtained in the intra-voxel structure estimations at fiber crossings, bifurcations, kissings, etc, benefit brain research, allowing to obtain better tractography estimations, hence, it results in an accurate computation of the brain connectivity patterns.

Ken Sakaie¹, Lael Stone¹, and Lowe Mark^1
1The Cleveland Clinic, Cleveland, OH, United States

Tractography-defined white matter pathways provide a means for improving the sensitivity of imaging to injury associated with specific function such as hand use or working memory. Natural variability along pathways may obscure differences associated with disease. We report on an anomaly in tissue integrity measures at the midline of the corpus callosum that may correspond to observations from histology.

Diffusion Kurtosis

Thursday 4 June 2015

Minxiong Zhou^1,2, Xu Yan³, and Guang Yang^2
1Shanghai Medical Instrumentation College, University of Shanghai for Science and Technology, Shanghai, China, ²Key Laboratory of Magnetic Resonance, East China Normal University, Shanghai, China, ³MR Collaboration NE Asia, Siemens Healthcare, Shanghai, China

The study applied vector-based nonlocal mean (VNLM) filter to diffusion kurtosis imaging (DKI). The VNLM filter considered multiple-b-value data, namely the diffusion decay curve, as a whole unit in similarity calculation, which is more robust than independently filtering each pixel in the decay curve. The results showed that the VNLM filter improve the visibility of tissue heterogeneity in tumor region, which provided sharper tissue structure and well suppressed background noise.

Akira Nishikori^1,2, Kohei Tsuruta^1,2, Koji Kamagata², Taku Hatano², Fumi Okuzumi², Masaaki Hori², Michimasa Suzuki², Shigeki Aoki², and Atsushi Seno^1
1Tokyo Metropolitan University, Arakawa-ku, Tokyo, Japan, ²Juntendo University School of Medicine, Bunkyo-ku, Tokyo, Japan

There are clinical proofs that the different clinical subtypes of Parkinson�fs disease (PD) have a different clinical course. Jellinger depicted a model of different projections of nigral dopaminergic neurons to striatal structures for the PD subtype. The purpose of our study is to detect and clarify microstructural changes of nigra-striatum dopaminergic neurons between akinetic-rigid and tremor-dominant Parkinson�fs disease by using high resolution diffusional kurtosis imaging (DKI) analysis. Mean kurtosis value in the contralateral posterior putamen were significantly higher in patients with akinetic-rigid type than in patients with tremor-dominant type, which is consistent with neuropathological model were depicted by Jellinger.

Andrey Chuhutin¹, Ahmad Raza Khan¹, Brian Hansen¹, and Sune Nørhøj Jespersen^1,2
1Center of Functionally Integrative Neuroscience, Aarhus University, Aarhus, Denmark, ²Dept. of Physics and Astronomy, Aarhus University, Denmark

As significant sensitivity of DKI to tissue pathologies was recently reported, this imaging modality is widely suspected to be useful for the neural tissue structure estimation. Being a Taylor expansion of the diffusion signal logarithm it deemed to have a validity region. We assessed the validity of the kurtosis evaluation versus for a range of b-values versus an analytically calculated ground truth based on the robust neurite model, that proved to finely describe the properties of the tissue. The results of the comparison show extremely high absolute error even for moderately small b-values both for extremely high and realistic SNR values.

Ben Jeurissen¹, Jan Sijbers¹, and Jacques-Donald Tournier^2,3
1iMinds-Vision Lab, Dept. of Physics, University of Antwerp, Antwerp, Belgium, ²Centre for the Developing Brain, King's College London, London, United Kingdom, ³Dept. of Biomedical Engineering, King's College London, London, United Kingdom

A crucial step in spherical deconvolution of diffusion-weighted MRI images is the definition of the single fiber response function. On one hand, advanced methods have been proposed to estimate per-subject response functions from the data. On the other hand, there is anecdotal evidence that these responses are relatively stable across subjects, advocating the use of a canonical response. In this study we investigate the inter-subject variability of the single fiber response from a large collection of data sets from unrelated healthy adult volunteers. Our findings suggest that in healthy volunteers the inter-subject variability is low, supporting the canonical response hypothesis.

Wen-Chau Wu^1,2, Han-Min Tseng³, and Ya-Fang Chen^4
1Graduate Institute of Oncology, National Taiwan University, Taipei, Taiwan, ²Graduate Institute of Clinical Medicine, National Taiwan University, Taipei, Taiwan, ³Department of Neurology, National Taiwan University Hospital, Taipei, Taiwan, ⁴Department of Medical Imaging, National Taiwan University Hospital, Taipei, Taiwan

We described a method for simultaneous measurement of cerebral blood volume and diffusion heterogeneity by combining diffusion kurtosis imaging with a two-compartment model. The derived diffusion kurtosis coefficient (K), diffusion coefficient (D), and blood volume fraction (f) were assessed for accuracy and precision at varied levels of signal-to-noise ratio (SNR) by computer simulations. Measurement precision was also assessed by bootstrap based on experimental data acquired from 15 healthy adult volunteers. Results showed that precision increases with SNR and that with a minimum baseline SNR of 64, coefficient of variation is ~10% for K, ~3% for D, and ~30% for f.

Elizabeth Zakszewski¹, Nathan Skinner², Shekar Kurpad¹, Brian Schmit³, and Matthew Budde^1
1Neurosurgery, Medical College of Wisconsin, Milwaukee, Wisconsin, United States, ²Biophysics, Medical College of Wisconsin, Milwaukee, Wisconsin, United States, ³Biomedical Engineering, Marquette University, Milwaukee, Wisconsin, United States

Noninvasive evaluation of the rodent spinal cord will be valuable to identify potential therapies for spinal cord injury and disease. We present improvements in multi-shot diffusion weighted EPI of the rat spinal cord for measurement of diffusion tensor and diffusion kurtosis parameters.

Austin Ouyang¹, Mihovil Pletikos², Nenad Sestan², and Hao Huang^1
1Advanced Imaging Research Center, University of Texas Southwestern Medical Center, Dallas, Texas, United States, ²Department of Neurobiology, Yale University, CT, United States

The primate cerebral cortex is characterized with complicated cytoarchitecture including neurons, glial cells, dendrites and small axons. Diffusion kurtosis imaging (DKI) has the potential to delineate the cortical microstructural complexity and provide complementary microstructural information to diffusion tensor imaging (DTI). Integrating cortical MK and FA map from high resolution DKI and DTI could offer a refreshing insight into the cellular microstructure noninvasively across the cerebral cortex. In this study, we aimed to reveal the MK and FA profile across different cortical areas and explore the relationship of cortical MK and FA with high resolution DKI and DTI of macaque brains.

Diffusion - Tractography

Thursday 4 June 2015

Helen Schomburg¹, Thorsten Hohage¹, Christoph Rügge¹, Sabine Hofer^2,3, and Jens Frahm^2
1Institute for Numerical and Applied Mathematics, Georg-August-Universität Göttingen, Göttingen, Germany, ²Biomedizinische NMR Forschungs GmbH, Max-Planck-Institut für biophysikalische Chemie, Göttingen, Germany, ³Bernstein Center for Computational Neuroscience, Göttingen, Germany

We present an algorithm for probabilistic tractography on HARDI data that exploits diffusion information of neighboring regions. In each iteration step, a guiding direction is determined from the previously obtained fiber fragment. Moreover, the region ahead is explored by computing a set of candidate directions and corresponding weights. This procedure is repeated recursively. The first set of candidate directions is assigned a probability based upon the final weight configuration. Then, a sample from this set is chosen randomly and contributes to a new tracking direction. The method is tested on a diffusion phantom as well as on in vivo data.

Haiyong Wu¹, Dinggang Shen¹, and Pew-Thian Yap^1
1Department of Radiology, University of North Carolina, Chapel Hill, North Carolina, United States

Global tractography algorithms require long computation times that are often prohibitive in the clinical setting. We propose a parallel implementation of global tractography that can take advantage of parallel computing technologies. Our method breaks the posterior density function of the configuration of fiber segments into a number of independent subposterior density functions, from which parallel sampling can be performed. Our results indicate that tractography results similar to the original global tractography algorithm proposed by Reisert et al. can be achieved using our method in a significantly reduced amount of time.

Etienne St-Onge¹, Gabriel Girard¹, Kevin Whittingstall², and Maxime Descoteaux^1
1Sherbrooke Connectivity Imaging Lab, Université de Sherbrooke, Sherbrooke, Québec, Canada, ²Department of Diagnostic Radiology, Faculty of Medicine and Health Science, Université de Sherbrooke, Sherbrooke, Québec, Canada

Conventional diffusion MRI fiber tracking methods have difficulty reconstructing white matter fiber structures near gray matter and have trouble penetrating into gyri and fully exploring the gyrus. Hence, the main limitation of current tractography techniques is the partial volume effect due to the tracking mask discretization in gyri and poor spatial resolution of the data. Our proposed surface tracking method successfully reconstructs fiber structures near the brain surface by only using a basic T1 anatomical image and shape information. With this method, we can see fanning in gyri, without the need of high resolution dMRI techniques.

Greg Parker¹, Mark Postans¹, and Derek Jones^1
1CUBRIC, School of Psychology, Cardiff University, Cardiff, South Glamorgan, United Kingdom

Tract-wise statistics are widely used within neuroscience research. Such measures may either be retrieved on a voxel-wise basis or, alternatively, through sampling along representative streamlines. We propose improvements to an existing shape-based streamline segmentation method to provide fully automatic segmentation of whole volume tractography datasets and, as evidence of this functionality, show that the method can be used to replicate the majority of findings from a recently published study.

Peter Savadjiev¹, Carl-Fredrik Westin², and Yogesh Rathi^1
1Psychiatry Neuroimaging Laboratory, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, United States, ²Laboratory for Mathematics in Imaging, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, United States

Graph theoretical representations of brain networks model the organization of gray matter units. We introduce a novel Dual graph formalism, in which the role of edges and vertices is inverted relative to the original (Primal) graph. This transformation shifts the emphasis of brain network analysis from gray matter units to their underlying connections. It applies standard graph theoretical operations to discover the organization of connections, as opposed to gray matter centers. Furthermore, it facilitates the characterization of each connection by a vector of several features. This is one solution to the problem of vector weights in standard brain network analysis.

Flavio Dell'Acqua¹, Luis Lacerda¹, Rachel Barrett¹, Lucio D'Anna², Stella Tsermentseli³, Laura Goldstein⁴, and Marco Catani^2
1Dept of Neuroimaging, King's College London, London, United Kingdom, ²Dept of Forensic and Neurodevelopmental Sciences, King's College London, London, United Kingdom,³Dept of Psychology, University of Greenwich, London, United Kingdom, ⁴Dept of Psychology, King's College London, United Kingdom

While manual dissections of tractography datasets may offer the best results in terms of anatomical accuracy, they are also extremely time consuming making difficult to extend them to large-scale datasets. On the contrary, automatic dissections or clustering approaches allow researcher to efficiently dissect large numbers of datasets but at the expense of decreased accuracy in the final dissection, leaving often little to no user interaction to control for artifactual components. In this study we propose a novel approach that drastically reduces the time required for manual dissections while preserving the ability to extract automatically tract specific measures from large datasets

Marc-Alexandre Côté¹, Eleftherios Garyfallidis¹, Hugo Larochelle¹, and Maxime Descoteaux^1
1Université de Sherbrooke, Sherbrooke, Québec, Canada

Tracking algorithms often generate non-reproducible streamlines which often appear to be anatomical outliers. Those streamlines affect connectivity studies. We propose a probabilistic outliers removal method based on a randomized hierarchical utilization of QuickBundles. Our method offers a way to reduce the number of invalid bundles while keeping reproducible valid bundles. This allows for more robust connectivity analyses.

Shu-Hsien Chu¹, Keshab K. Parhi¹, and Christophe Lenglet^1
1University of Minnesota, Minneapolis, Minnesota, United States

In the paper, a novel brain network is proposed with nodes as brain regions, links as possible white matter fiber bundles, flow as electrochemical signal, link capacities characterized by fiber strength based on diffusion MRI, and node demands as neural reaction estimated from functional MRI. The signaling pathways are discovered through solving the proposed brain network model. Comparing with the connectivity derived from either diffusion MRI, functional MRI, or a joint model using the expectation-maximization algorithm presented in a prior work, the proposed model finds the maximum true connections with fewest number of false connections.

Lea Vinokur^1,2, Andrew Zalesky^3,4, David Raffelt¹, Robert Smith¹, and Alan Connelly^1,2
1The Florey Institute of Neuroscience and Mental Health, Heidelberg, Victoria, Australia, ²Department of Florey Neurosciences, University of Melbourne, Melbourne, Victoria, Australia, ³Melbourne School of Engineering, University of Melbourne, Melbourne, Victoria, Australia, ⁴Melbourne Neuropsychiatry Centre, University of Melbourne, Melbourne, Victoria, Australia

The connectome is becoming an increasingly popular tool to study brain connectivity. Case-control study at the level of individual connections is difficult due to a multiple comparisons problem. We propose a new method to combine Network Based Statistics, a statistical framework developed to adapt cluster-based inference to a network, with TFCE, a method to boost belief in signal clusters and remove the dependence on arbitrary thresholds. We apply the combined framework, denoted "NBS-TFCE", to in vivo structural connectivity data with synthetically introduced pathologies, to try to determine optimal parameters for performing NBS-TFCE on realistic connectivity matrices."

Christian Wieseotte^1,2, Thomas Witzel³, Jon Polimeni³, Aapo Nummenmaa³, Bernhard Gruber⁴, Laura Schreiber^1,5, and Lawrence Wald^3
1Department of Radiology, Section of Medical Physics, Johannes Gutenberg University Medical Center, Mainz, Germany, ²Max Planck Graduate Center, Mainz, Germany,³Department of Radiology, Massachusetts General Hospital, Athinoula A. Martinos Center for Biomedical Imaging, Boston, MA, United States, ⁴Department for Medical Engineering, University of Applied Sciences Upper Austria, Linz, Austria, ⁵Department of Cellular and Molecular Imaging, Comprehensive Heart Failure Center, Würzburg, Germany

Because of long measurement times and the absence of motion during image acquisition, ex-vivo DWI is capable of achieving significantly higher spatial and angular resolutions compared to routine in-vivo imaging. With smaller voxel volumes however, SNR becomes the limiting factor for increasing the spatial resolution. The purpose of this study was to explore these limits in ex-vivo DWI and tractography by maximizing SNR. With a 60 channel coil array optimized for post mortem human brain specimen, a fast segmented 3D EPI acquisition and powerful 300 mT/m gradients, an isotropic resolution of 350µm was achieved.

Francois Rheault¹, Jean-Christophe Houde¹, and Maxime Descoteaux^1
1Université de Sherbrooke, Sherbrooke, Quebec, Canada

With the current, fast-paced advances in imaging and tractography techniques, streamlines files are poised to become incredibly massive. With this increased size comes challenges for real-time visualization and selection of streamlines. To be able to overcome those challenges, we propose a new load-time simplification mechanism, which allows loading and interactively working with files containing millions of streamlines without any visual artefact. This mechanism also implies updating the fiber selection mechanisms, to make sure that no streamline goes missing. We show that our updated selection technique allows interactive selection on huge datasets, while still providing exact selection results.

G. Russell Glenn¹, Jens H. Jensen², Yi-Ping Chao³, Chu-Yu Lee², Joseph A. Helpern⁴, and Li-Wei Kuo^5
1Neurosciences & Center for Biomedical Imaging, Medical Univesity of South Carolina, Charleston, SC, United States, ²Radiology & Center for Biomedical Imaging, Medical Univesity of South Carolina, SC, United States, ³Computer Science and Information Engineering, Chang Gung University, Taoyuan, Taiwan, ⁴Radiology, Neurosciences, & Center for Biomedical Imaging, Medical Univesity of South Carolina, SC, United States, ⁵Institute of Biomedical Engineering and Nanomedicine, National Health Research Institutes, Miaoli County, Taiwan

This study compares the diffusion orientation distribution function (dODF) approximations from diffusional kurtosis imaging (DKI) and diffusion tensor imaging (DTI) to the dODF reconstructions from diffusion spectrum imaging (DSI). The DKI approximation of the dODF can resolve intra-voxel white matter (WM) fiber crossings for WM fiber tractography (FT) with comparable angular accuracy to DSI, which improves upon DTI based WM fiber tractography (FT). With lower scanning requirements, DKI may be more suitable for clinical applications than DSI. DKI-based WM FT and associated quantitative indices may help improve our understanding of neural connectivity in normal and pathological states.

Chun-Hung Yeh¹, Robert Smith¹, Xiaoyun Liang¹, Fernando Calamante^1,2, and Alan Connelly^1,2
1The Florey Institute of Neuroscience and Mental Health, Heidelberg, Victoria, Australia, ²Department of Medicine, Austin Health and Northern Health, University of Melbourne, Melbourne, Victoria, Australia

Quantification of structural connectomics based on streamline connection density is known to be problematic due to the fundamental limitations of diffusion MRI tractography. This study demonstrates how common brain network metrics of structural connectomes vary considerably when novel streamline reconstruction techniques such as ACT (anatomically-constrained tractography) and SIFT (spherical-deconvolution informed filtering of tractograms) are applied, and that a popular heuristic correction for streamline connection density based on length scaling is not an adequate substitute for these methods, highlighting the necessity for the use of such advanced reconstruction techniques to provide connectome construction with robust and accurate quantitative properties.

Esha Datta¹, Kesshi Jordan¹, Eduardo Caverzasi¹, and Roland Henry^1
1University of California, San Francisco, San Francisco, California, United States

Diffusion MRI tractography if often used in pre-neurosurgical planning to map brain connections that are considered critical to motor, visual, and language function. Usually, this data is segmented manually through a time consuming process requiring a trained technician. This study explores the use of an alternative automatic classification method, which uses a training set to output a set of classified tracts from a set of streamlines. This method correctly identifies the rough volume of all tracts tested and for the left IFOF, the tracts classified by the algorithm and the tracts classified by humans were almost indistinguishable (P-value = .9021).

YISHAN LUO¹, LIN SHI^2,3, WINNIE CW CHU¹, VINCENT CT MOK², and Defeng Wang^1,4
1Dept of Imaging and Interventional Radiology, The Chinese University of Hong Kong, Hong Kong, Hong Kong, Hong Kong, ²Dept of Medicine and Therapeutics, The Chinese University of Hong Kong, Hong Kong, Hong Kong, ³Chow Yuk Ho Technology Centre for Innovative Medicine, The Chinese University of Hong Kong, Hong Kong, Hong Kong,⁴Department of Biomedical Engineering and Shun Hing Institute of Advanced Engineering, The Chinese University of Hong Kong, Hong Kong, Hong Kong

In this paper, we propose a feature-based tractography image registration method, targeting at achieving good tract alignment for tract-level DTI analysis across subjects. Although some DTI registration methods can well match tensor images, most of them fail to achieve good tract alignment. Our method first uses an atlas-based WM paracellation method to establish bundle-to-bundle correspondence. Then with three features designed for each fiber, fiber-to-fiber correspondence is set up for tractogrpahy image registration. Experimental results validated that our registration can greatly outperform DTI registration in terms of tract alignment, which demonstrates that our method can help realize more accurate tract-level analysis.

Eleftherios Garyfallidis¹, Marc-Alex Côté¹, Janice Hau², Guy Perchey², Laurent Petit², Stephen C. Cunnanne³, and Maxime Descoteaux^1
1Département d’informatique, Faculté des Sciences, Université de Sherbrooke, Sherbrooke, Quebec, Canada, ²GIN UMR5296 CNRS CEA, Université de Bordeaux, France,³Research Center on Aging and Department of Medicine, Université de Sherbrooke, Quebec, Canada

We introduce a novel method for efficient and accurate automatic recognition of white matter bundles using prior models and local streamline-based optimization.

Gabriel Girard^1,2, Kevin Whittingstall³, Rachid Deriche², and Maxime Descoteaux^1
1Sherbrooke Connectivity Imaging Lab (SCIL), Université de Sherbrooke, Sherbrooke, Quebec, Canada, ²Project Team Athena - INRIA, Sophia Antipolis, France, ³Department of Diagnostic Radiology, Faculty of Medicine and Health Science, Université de Sherbrooke, Sherbrooke, Quebec, Canada

In this study, we investigate the reproducibility of the connectivity matrix, resulting from different tractography parameters. We vary the number of streamlines used to construct the matrix in cortical to cortical connectivity and analyze its effects. We show that the reproducibility of the connectivity is surprisingly similar across tractography pipeline. In all cases, connectivity matrices tend to stabilize using more than 100,000 streamlines. We found that the analysis of the reproducibility of tractography is a first step to find which tractography pipeline is more characteristic of the underlying anatomical structure.

Anneriet M Heemskerk¹, Michel Thiebaut de Schotten², Marco Catani², Silvio Sarubbo³, Laurent Petit⁴, Max Viergever¹, Derek K. Jones⁵, John Evans⁵, Tomáš Paus^6,7, and Alexander Leemans^1
1Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands, ²King's College London, United Kingdom, ³Santa Chiara Hospital, Italy, ⁴GIN-UMR5296, CNRS, CEA, University of Bordeaux, Bordeaux, France, ⁵Cardiff University, United Kingdom, ⁶Rotman Research institute, Baycrest, Toronto, Canada, ⁷Departments of Psychology and Psychiatry, University of Toronto, Toronto, Canada

Diffusion-weighted MRI based fiber tractography (FT) is widely used and its methodology is constantly improving. Despite the many pitfalls and limitations of FT, we present a description of fiber tract pathways in the orbitofrontal / prefrontal cortex, which – to the best of our knowledge – have not been reported before with FT. A consistent tract configuration across multiple subjects and different data cohorts was observed which boost our confidence that this finding is not based on artifacts.

Sofya Kulikova¹, Jessica Dubois², Pamela Guevara³, Jean-François Mangin⁴, Catherine Chiron⁵, Nicole Chemaly⁵, Silvia Napuri⁶, Cyril Poupon⁷, and Lucie Hertz-Pannier^1
1INSERM UMR1129, CEA/Neurospin/UNIACT, Université Paris Descartes, Sorbonne Paris Cité, Paris, France, ²INSERM UMR992, CEA/Neurospin/UNICOG, Université Paris Sud, Paris, France, ³University of Concepción/Departamento de Ingeniería Eléctrica, Chile, ⁴CEA/Neurospin/UNATI, Gif-sur-Yvette, France, ⁵INSERM UMR1129, Université Paris Descartes, Sorbonne Paris Cité, Paris, France, ⁶Pediatric Department, CHU Hôpital Sud, Rennes, France, ⁷CEA/Neurospin/UNIRS, Gif-sur-Yvette, France

Tractography datasets are extremely complex and extracting individual bundles from them is still a challenging task. Recently, fiber-clustering techniques that take into account fiber shapes and localization variabilities were proposed for automatic bundles identification, based on an atlas of main bundles. However, this atlas was generated for adults hindering its application to children as fiber shapes and lengths change during development. In this work we present a child brain connectivity atlas, which can be used in studies on normal and pathological brain development for automatic bundles identification and further evaluation of the MRI parameters across the identified bundles.

David Raffelt¹, Robert E Smith¹, J-Donald Tournier^2,3, Gerard R Ridgway^4,5, David Vaughan^1,6, and Alan Connelly^1,7
1Florey Institute of Neuroscience and Mental Health, Melbourne, VIC, Australia, ²Centre for the Developing Brain, King's College London, London, United Kingdom, ³Department of Biomedical Engineering, King's College London, London, United Kingdom, ⁴FMRIB Centre, University of Oxford, Oxford, United Kingdom, ⁵UCL Institute of Neurology, University College London, London, United Kingdom, ⁶Department of Medicine, University of Melbourne, Melbourne, Australia, ⁷The Department of Florey Neuroscience and Mental Health, University of Melbourne, Melbourne, VIC, Australia

Voxel-based analysis is being increasingly used to study white matter development, aging and pathology. Connectivity-based Fixel Enhancement (CFE) is a recently developed statistical method that enables whole-brain analysis of fibre-specific diffusion MRI measures within regions containing crossing fibres. While this method does not require an arbitrary test-statistic threshold, it is dependent on other parameters for the enhancement step. We assessed CFE performance by introducing simulated pathology into in vivo data, and explored combinations of enhancement parameters while varying the pathology region, effect size and pre-smoothing spatial extent. Results suggest CFE parameters are relatively insensitive to pathology region and effect size.

Joanes Grandjean¹, Zsófia Pröhle², and Markus Rudin^1,3
1Institute for Biomedical Engineering, ETH and University Zurich, Zurich, Switzerland, ²Department of Physics, ETH Zurich, Zurich, Zurich, Switzerland, ³Institute of Pharmacology and Toxicology, University Zurich, Zurich, Switzerland

Resting-state fMRI networks are generally organized with a bilateral structure in humans and rodents. Yet, a number of these networks are not supported by direct synaptic connections. This is the case for the caudate-putamen, which is part of the dorsal striatum network in mice. An intermediately region is assumed to relay the information between the two regions; however, such a region is not apparent on the functional connectivity maps. We first compared DTI based tractography in mice with injection-based tractography available from the Allen institute database. We then used probabilistic tracking to identify the fibers involved between each functional network.

Yuguang Meng¹ and Xiaodong Zhang^1,2
1Yerkes Imaging Center, Yerkes National Primate Research Center, Emory University, Atlanta, GA, United States, ²Division of Neuropharmacology and Neurologic Diseases, Yerkes National Primate Research Center, Emory University, Atlanta, GA, United States

The knowledge of the interhemispheric connectivity alterations in non-human primates during development and aging may provide important implications for functional and behavioural alterations in human. Compare to diffusion tensor imaging, diffusion spectrum imaging (DSI) tractography provides a novel and unprecedented opportunity for studying complicated fiber connections. In this work, DSI tractography was used to evaluate the cortically specific changes of transcallosal connectivity of formalin-fixed macaque brains from infancy to late adulthood, and differential change patterns were found in anterior to posterior brain lobes.

Tyler Rehbein¹, Michelle Moerel², Frederico De Martino³, An Vu², Essa Yacoub², and Christophe Lenglet^2
1University of Minnesota Medical School, Minneapolis, MN, United States, ²Center for Magnetic Resonance Research, Department of Radiology, University of Minnesota, Minneapolis, MN, United States, ³Department of Psychology and Neuroscience, Maastricht University, Maastricht, Netherlands

This study compares high-resolution 3 tesla (3T) and 7 tesla (7T) diffusion MRI data from the Human Connectome Project (HCP) using probabilistic tractography to look for reductions in spurious connections and improvements in central auditory pathway characterization. Auditory cortex identified using a probabilistic atlas in 2 subjects used as a seed along with manually segmented waypoints in the medial geniculate nucleus and inferior colliculus successfully characterized central auditory pathways. The auditory radiation and colliculogeniculate pathways were identified with 3T and 7T diffusion data, and subcollicular pathways were also successfully traced with 7T diffusion data.

Gwendolyn Van Steenkiste¹, Julie Hamaide², Ben Jeurissen¹, Dirk H.J. Poot^3,4, Johan Van Audekerke², Jan Sijbers¹, and Marleen Verhoye^2
1iMinds-Vision Lab, University of Antwerp, Antwerp (Wilrijk), Antwerp, Belgium, ²Bio-Imaging Lab, University of Antwerp, Antwerp, Belgium, ³BIGR (Medical informatics and Radiology), Erasmus Medical Center Rotterdam, Rotterdam, Netherlands, ⁴Imaging Science and Technology, Delft University of Technology, Delft, Netherlands

Histology does not allow quantitative investigation of structural connectivity on a whole-brain level, in contrast to MRI techniques such as Diffusion Tensor Imaging (DTI). Recently, an in vivo DTI study revealed a novel sexual dimorphism in the zebra finch brain. In order to achieve a precise understanding of the anatomical substrate underlying the observed difference, we propose to use super-resolution DTI (SR-DTI) and track density imaging (TDI). We show that by combining SR-DTI and TDI, a clear anatomical contrast of song control system connectivity as well as a clear delineation of its different components can be realized.

Diffusion Outside the Brain

Thursday 4 June 2015

Fabian Hilbert¹, Simon Veldhoen¹, Tobias Wech¹, Henning Neubauer¹, Thorsten Bley¹, and Herbert Köstler^1
1Departement of Radiology, University of Würzburg, Würzburg, Germany

The commonly used model of Intravoxel incoherent motion (IVIM) implies an isotropic perfusion fraction f. Diffusion tensor imaging (DTI) describes anisotropy in the diffusion. A simple combination of IVIM and DTI still cannot distinguish isotropic from anisotropic perfusion fraction. We propose a perfusion fraction tensor model in combination with IVIM-DTI. The presented model is compared to an isotropic perfusion model in the kidney using Akaike’s information criterion. Six healthy volunteers have been examined. In large part cortex and medulla are described equally or better by the perfusion fraction tensor model than by the isotropic perfusion fraction model.

Dariya Malyarenko¹, Yuxi Pang¹, Julien Senegas², Marko Ivancevic³, Brian D Ross¹, and Thomas L Chenevert^1
1Radiology, University of Michigan, Ann Arbor, Michigan, United States, ²Philips Research Laboratories, Hamburg, Germany, ³Philips Healthcare, Best, Netherlands

Nonuniform diffusion weighting bias due to gradient nonlinearity causes substantial errors in ADC maps for anatomical regions imaged distant from magnet isocenter. Our previously-described approach allowed effective removal of spatial ADC bias for mono-exponential media of arbitrary anisotropy. Here we evaluate correction performance for quantitative diffusion parameters of the IVIM model for renal tissue. Comparable accuracy is achieved both for corrections based on b-maps and DWI intensities for “slow” ADC component in presence of IVIM.

Matthew R Orton¹, Neil Peter Jerome¹, Evangelia Kaza¹, David J Collins¹, Dow-Mu Koh², Bernd Kuehn³, and Martin O Leach^1
1Radiotherapy and Imaging Department, Institute of Cancer Research, Sutton, Surrey, United Kingdom, ²Department of Radiology, Royal Marsden Hospital, Sutton, Surrey, United Kingdom, ³Siemens Medical Solutions, Erlangen, Germany

Imaging in the abdomen and thorax is more challenging than other regions due to various forms of tissue motion. For diffusion weighted imaging, registration techniques need to be robust to the wide range of image brightness and contrast over the b-values typically used. This abstract presents an image registration algorithm for diffusion weighted images that uses a multi-exponential model of signal attenuation to sequentially synthesize target images with matching brightness and contrast on which to register the source images. This technique is highly effective for removing in-plane motion with coronally acquired images.

Tong San Koh¹, Septian Hartono¹, Tiffany P. Hennedige¹, Yet Yen Yan¹, In Chin Song², Lin Zheng², Wing Sum Lee², Helmut Rumpel³, Laurent Martarello⁴, James B.K. Khoo¹, Dow-Mu Koh⁵, and Choon Hua Thng^1
1National Cancer Centre Singapore, Singapore, Singapore, ²SingHealth Experimental Medicine Centre, Singapore, Singapore, ³Singapore General Hospital, Singapore, Singapore, ⁴Roche-Singapore Translational Medicine Hub, Singapore, Singapore, ⁵Royal Marsden Hospital, Surrey, United Kingdom

The purpose of this study was to explore possible alterations in intravoxel incoherent motion (IVIM) model diffusion and perfusion parameters with the development of fibrosis in the kidney, using a murine model of unilateral ureteral obstruction (UUO). IVIM analysis revealed a decrease in both D and f in the renal parenchyma with the development of fibrosis, and suggested possible microvascular contribution to the reduction in ADC.

Scott A. Willis¹, Timothy Stait-Gardner¹, William S. Price¹, and Roger Bourne^2
1Nanoscale Organisation and Dynamics Group, School of Science and Health, University of Western Sydney, Sydney, NSW, Australia, ²Discipline of Medical Radiation Sciences, Faculty of Health Sciences, University of Sydney, Sydney, NSW, Australia

Self-diffusion reports on binding events, restrictions and anisotropies of the system from porous rock to biological tissue, and can be measured using pulsed gradient spin-echo (PGSE) NMR. Double-pulsed gradient spin-echo (DPGSE) is useful for probing molecular dynamics as well as pore sizes and orientation, giving more information than single PGSE measurements. This work investigates angular DPGSE profiles (i.e., equivalent gradient strengths for each PGSE block of the DPGSE but with varying angular separation) in prostate tissue and the use of DTI results to simulate the DPGSE profile for a ‘one-voxel’ DPGSE measurement.

Sisi Liang¹, Eleftheria Panagiotaki², Peng Shi¹, and Roger Bourne^3
1College of Engineering and Science, Victoria University, Melbourne, Vic, Australia, ²Centre for Medical Image Computing, University College London, London, England, United Kingdom, ³Discipline of Medical Radiation Sciences, Faculty of Health Sciences, University of Sydney, Sydney, NSW, Australia

Structure-based modelling of DWI in tissue is becoming a powerful tool for probing microstructure and pathology. A recent study demonstrated the feasibility of a three-component structure-based model called VERDICT for imaging prostate cancer in vivo with DWI at clinical setting with high sensitivity and specificity. Motivated by this new finding, this work investigated the non-perfusion components of the VERDICT model in prostate tissue ex vivo at 9.4 T. Seven compartment models with up to 3 components were compared using Akaike¡¯s information criterion. The best-performed model had two compartments with one anisotropic and one restricted compartment.

Oliver Gurney-Champion^1,2, Martijn Froeling³, Remy Klaassen^4,5, Hanneke W.M. van Laarhoven⁴, Jaap Stoker¹, Arjan Bel², and Aart J. Nederveen^1
1Radiology, Academic Medical Center, Amsterdam, Netherlands, ²Radiation Oncology, Academic Medical Center, Amsterdam, Netherlands, ³Radiology, University Medical Center Utrecht, Utrecht, Netherlands, ⁴Department of Medical Oncology, Academic Medical Center, Amsterdam, Netherlands, ⁵Laboratory for Experimental Oncology and Radiobiology, Academic Medical Center, Amsterdam, Netherlands

Measurements suitable for intra-voxel incoherent motion (IVIM) modelling of diffusion weighted (DW) MRI take long (>10 minutes) as multiple averages and diffusion weightings are measured. To determine the minimal measurement time, we investigated how IVIM in the liver and pancreas performs as function of measurement time. We measured DW-MRI (9 averages, 14 b-values) in 16 healthy volunteers twice in two sessions. During post processing we deleted measurements to assess the performance as function of averages and b-values. We find that IVIM can be done roboustly in 5 minutes using 5 averages and 12 b-values.

Ryan Pathak¹, Hossein Ragheb², Neil A Thacker², David Morris², and Alan Jackson^1
1The Wolfson Molecular Imaging Centre, University of Manchester, Manchester, United Kingdom, ²Centre for Imaging Sciences, University of Manchester, Manchester, United Kingdom

ADC is a potential biomarker of cell death, and an early indicator of treatment success or failure in oncology. Reported ADC reproducibility has been poor. This is a multi-site and vendor study in liver metastasis from colorectal tumours. We compute variation of mean ADC between sites. We developed a statistical model of expected errors to highlight the relative importance of factors affecting reproducibility in an individual. We use this model to effectively demonstrate that reproducibility of less than 4% is achievable with optimisation of factors, such as region size, that influence the variation in ADC for an individual.

Sila Kurugol¹, Moti Freiman¹, Onur Afacan¹, Sean Clancy¹, and Simon K Warfield^1
1Radiology, Boston Children's Hospital and Harvard Medical School, Boston, Massachusetts, United States

Diffusion-weighted MRI enables characterization of abnormalities including liver fibrosis and tumors through measurement of variations in the mobility of water molecules. The Intra-voxel incoherent motion (IVIM) model represents the diffusion signal decay with a bi-exponential function with one decay rate parameter for slow, and a second decay rate parameter for fast diffusion associated with microcirculation. Recently, a spatially-constrained probability distribution model (SPIM) was introduced to represent the heterogeneity of the diffusion components with a two-component mixture model with a spatial homogeneity prior. Here, we evaluate the longitudinal reproducibility of parameters estimation using SPIM compared to IVIM in 68 abdominal scans.

Aritrick Chatterjee¹, Geoff Watson², Esther Myint³, Paul Sved², Mark McEntee¹, and Roger Bourne^1
1Faculty of Health Sciences, University of Sydney, Sydney, New South Wales, Australia, ²Royal Prince Alfred Hospital, Sydney, New South Wales, Australia, ³Douglass Hanly Moir Pathology, Sydney, New South Wales, Australia

This study investigates the biophysical basis of clinically observed decrease of ADC in prostate cancer. ADC predicted from gland component partial volumes correlated strongly with measured ADC in both fresh and fixed tissue. Epithelium and lumen partial volumes each correlated more strongly with measured ADC than ‘cellularity’ metrics: nuclear count and nuclear area. Differences in the partial volume of prostate gland components having distinct diffusivities, rather than changes in the conventionally cited ‘cellularity’ metrics, are likely to be the major contributor to clinically observed variations of ADC in prostate tissue.

Eleftheria Panagiotaki¹, Andrada Ianus¹, Edward Johnston², Rachel W Chan², Nicola Stevens², David Atkinson², Shonit Punwani², David J Hawkes¹, and Daniel C Alexander^1
1Centre for Medical Image Computing, University College London, London, London, United Kingdom, ²Centre for Medical Imaging, University College London, London, United Kingdom

This method provides a clinically feasible imaging protocol of 10 minutes for prostate tissue characterization with VERDICT (Vascular Extracellular and Restricted DIffusion for Cytometry in tumours). Previous work used long diffusion imaging protocols in the border of what some patients are willing to tolerate. This work uses a computational optimization framework with VERDICT to meet the clinical scan duration that will enable larger clinical trials for widespread translation. Initial results on patients using the optimised protocol show promise for cancer delineation via the microstructural VERDICT estimates.

Feifei Qu¹, Ramkumar Krishnamurthy², Pei-Herng Hor^1,3, John Fisher⁴, Claudio Arena⁴, Debra Dees⁴, and Raja Muthupillar^4
1Physics Department, University of Houston, Houston, TX, United States, ²Radiology Department, Texas Children's Hospital, Houston, TX, United States, ³Texas Center for Superconductivity, Houston, TX, United States, ⁴Diagnostic and Interventional Radiology, St. Luke's Medical Center, Houston, TX, United States

In recent study, f map of intravoxel incoherent motion(IVIM) model which indicates the blood volume ratio was applied to measure the treated region of uterine fibroid in magnetic resonance guided high focused ultrasound treatment. In this study, it was reported that the T2 value increased in the treated region affected the asymptotic fitting for IVIM. The mismatch of the perfusion information and thermal dose for the uterine fibroid after HIFU therapy can be eliminated by T2 correction.

Hao Fu¹, Chenxia Li¹, Rong Wang¹, Jianxin Guo¹, and Jian Yang^1
1Department of Radiology, the First Affiliated Hospital of Xi'an Jiaotong University, Xi'an, Shaanxi, China

The aim of the study is to investigate the variation among the different T2WI types uterine fibroids by analysis of ADC histogram. Eighty patients divided in type1 of 34 patients, type2 of 36 patients, type3 of 10 patients, respectively. Conventional MRI and DWI were performed. Then, ADCq, kurtosis and skewness were derived from ADC histogram. The results showed that values of ADCq and kurtosis were significant difference between type 1 and type2£¬type1 and type3. The kurtosis among three type are contrary to ADCq, which of type1 is lowest than others. However, there is no significant difference between type2 and type3 in ADCq, kurtosis and skewness. Therefore, ADCq and kurtosis can provide quantitative information to identify different fibroid types, which could be used as useful screening tools to guide the patients selection for MRgHIFU ablation.

Konrad Schieban¹, Timothy G Reese², Christian T Stoeck¹, David E Sosnovik², Sebastian Kozerke^1,3, and Choukri Mekkaoui^2
1Institute for Biomedical Engineering, ETH Zurich, Zurich, Switzerland, ²Radiology, Harvard Medical School, Massachusetts General Hospital, Martinos Center for Biomedical Imaging, Charlestown, MA, United States, ³Division of Imaging Sciences, King's College London, London, United Kingdom

Large gradient amplitudes in cardiac spin echo DTI are desirable, since they allow gradient pulses to be short while still providing appropriate diffusion sensitivity. However, it is unclear if the benefit of motion robustness is offset by larger eddy currents which are increasingly influential at high field strengths. In this work, the impact of motion on the diffusion tensor is analyzed at 80mT/m and 150mT/m and compared with the impact of eddy currents which have been measured on the Siemens MAGNETOM Skyra CONNECTOM. Furthermore, geometric distortions due to eddy currents are demonstrated in phantom experiments.

Haiyan Sun¹, Shaowu Wang², Ziheng Zhang³, Weisheng Zhang¹, Lina Zhang¹, Minting Zheng¹, Meiyu Sun¹, Qingwei Song¹, and Dianxiu Ning^1
1Radiology department, The first hospital affiliated to Dalian Medical University, Dalian, Liaoning, China, ²Radiology department, The second hospital affiliated to Dalian Medical University, Dalian, Liaoning, China, ³GE Healthcare China,Beijing, Beijing, China

Fat cell tumors constitute the largest group of the soft tissue tumors. lipomas account for at least 30% of benign soft tissue tumors. Liposarcoma are most common soft tissue sarcomas. Typical lipomas are easy to identified by conventional MRI, characterized with high signals on T1WI and T2WI, signal are reduced in fat suppression sequence. But atypical lipomas, including space, are difficult to diagnosis. Diffusion weighted imaging (DWI) can detect the signal strength according to the difference of water molecular diffusion movement.The apparent diffusion coefficient (ADC) values, as quantitative parameters, can be used in the evaluation of benign and malignant tumors.

Nina L. Purvis¹, Peter Gibbs², Martin D. Pickles², and Lindsay W. Turnbull^2
1Centre for MR Investigations, Hull York Medical School, Hull, East Yorkshire, United Kingdom, ²Centre for MR Investigations, University of Hull at HYMS, Hull, East Yorkshire, United Kingdom

A study to investigate the presence and repeatability of IVIM in breast parenchyma of healthy volunteers. Two optimised IVIM protocols were applied in 11 healthy volunteers, and the data was fitted mono- and bi-exponentially with a cut-off for D of 200s/mm2. RMSEs indicated that the biexponential fit was better. The repeatability of D for b-values10 and b-values20 was 9% and 20% respectively, with the repeatability of f and D* showing more variation at up to ±0.12 and ±0.039s/mm2. The results indicated that there was an IVIM effect in breast parenchyma.

Lin Cheng¹, Matthew D. Blackledge¹, David J. Collins¹, Nina Tunariu¹, Martin O. Leach¹, and Dow-Mu Koh^1
1Institute of Cancer Research, Sutton, London, United Kingdom

Diffusion weighted magnetic resonance (DW-MR) imaging in the body is used for tumour detection based on its high contrast between lesion and normal tissue. Previous studies have shown that computed diffusion-weighted MR Imaging (cDWI) provides a means of improving image contrast through synthesis of high-b-value DW images. The cDWI method has intrinsic T2 contrast, estimating T2 allows us to improve the cDWI technique by exploiting variable T2 contrast. The purpose of this study is to describe a modified cDWI model that provides synthetic images at arbitrary b-values and echo-times. We demonstrate the methods improved image contrast and tumour detection.