Eleftheria Panagiotaki¹, Torben Schneider^1,2, Bernard Siow^1,3, Mark F Lythgoe³, Matt G Hall¹, and Daniel C Alexander^1
1Centre for Medical Image Computing, Dept. of Computer Science, University College London, London, United Kingdom, ²Institute of Neurology, University College London, ³Centre for Advanced Biomedical Imaging, University College London

This study aims to identify the minimum requirements for an accurate model of the diffusion MR signal in white matter of the brain. We construct a hierarchy of multi-compartment models of white matter from combinations of simple models for the intra- and the extra-axonal spaces. We devise a new diffusion MRI protocol that provides measurements with a wide range of parameters for diffusion sensitization both parallel and perpendicular to white matter fibres. We use the protocol to acquire data from a fixed rat brain, which allows us to fit, study and compare the different models.

David Raffelt^1,2, J-Donald Tournier^3,4, Gerard Ridgway⁵, Stephen Rose⁶, Robert Henderson⁷, Stuart Crozier², Alan Connelly^3,4, and Olivier Salvado^1
1The Australian E-Health Research Centre, CSIRO, Brisbane, QLD, Australia, ²Biomedical Engineering, School of ITEE, University of Queensland, Brisbane, QLD, Australia, ³Brain Research Institute, Florey Neuroscience Institutes (Austin), Melbourne, VIC, Australia, ⁴Department of Medicine, University of Melbourne, Melbourne, VIC, Australia, ⁵Institute of Neurology, University College London, London, United Kingdom, ⁶Centre for Advanced Imaging, University of Queensland, Brisbane, QLD, Australia, ⁷Department of Neurology, Royal Brisbane and Women's Hospital, Brisbane, QLD, Australia

Apparent Fibre Density (AFD) is a new measure based on information provided by Fibre Orientation Distributions. AFD enables voxel-based analysis to be performed over space and orientation, and therefore population differences may be attributed to a single fibre within a voxel containing multiple fibres. Performing comparisons over many orientations within each voxel increases the number of multiple comparisons. We present a method for cluster-based inference of spatially extended differences in AFD by identifying clusters of contiguous supra-threshold directions using neighbours defined in space and orientation. The proposed method is demonstrated using a cohort of Motor Neurone Disease and healthy subjects.

Anh Tu Van¹, Rafael O'Halloran¹, Samantha Holdsworth¹, and Roland Bammer^1
1Radiology, Stanford University, Stanford, CA, United States

Despite its much richer information regarding the microstructure of neuronal architectures than other imaging techniques, such as diffusion tensor imaging (DTI) and high angular resolution diffusion imaging (HARDI), applications of diffusion spectrum imaging (DSI) in in vivo studies are limited due to its long total acquisition time. The current study proposes to significantly speed up DSI by using partially encoding in q-space, similar to partial Fourier encoding in traditional k-space domain. Both phantom and in vivo experiments are done to show the performance of the proposed method. For the experiments presented, the achieved acquisition speed-up was 44.4%.

Marion Irene Menzel¹, Jonathan Immanuel Sperl¹, Ek Tsoon Tan², Kedar Khare², Kevin F King³, Xiaodong Tao², Christopher J Hardy², and Luca Marinelli^2
1GE Global Research, Garching bei München, Germany, ²GE Global Research, Niskayuna, NY, United States, ³GE Healthcare, Waukesha, WI, United States

The combination of undersampled acquisition and reconstruction using compressed sensing enables the acceleration of diffusion spectrum imaging, bringing this application closer to clinical practice. This work evaluates the performance of compressed sensing as a function of data reconstruction size, pattern pa-rameter and specific realization of the random pattern. Among the sampling pattern distributions we tested both in simulations and in vivo data from brains of healthy volunteers, our results demonstrate that Gaussian undersampling performed best. Especially for higher acceleration factors and small matrix sizes the appropriate realization of the sampling pattern from the random distribution has to be chosen carefully.

Etienne Saint-Amant¹, and Maxime Descoteaux^1
1Computer Science Department, Université de Sherbrooke, Sherbrooke, Québec, Canada

In a Compressed Sensing (CS) framework, we focused on characterizing the sparsity of diffusion propagator from Diffusion Spectrum Imaging (DSI) data. We extensively analyzed the performance of the 3D orthogonal wavelet basis and biorthogonal wavelet basis. We answer important questions such as: What is the best basis? Is biorthogonality showing benefits over orthogonality? What is the best thresholding method? How does this apply in a real human brain dataset?

Greg Daniel Parker^1,2, Nicholas Drage^3,4, Paul L Rosin², A David Marshall², Stephen Richmond⁴, John Evans¹, and Derek K Jones^1
1CUBRIC, School of Psychology, Cardiff University, Cardiff, United Kingdom, ²School of Computer Science, Cardiff University, Cardiff, United Kingdom, ³Cardiff Vale NHS Trust, United Kingdom,⁴School of Dentistry, Cardiff University, United Kingdom

Accurate in vivo estimates of muscle fibre trajectory are desirable for evaluation of subject-specific maxillofacial surgical treatment options. While diffusion tensor MRI provides adequatly reconstructs larger muscles (e.g. calf), fibre crossing inherent to maxillofacial musculature exposes well-known limitations; necessitating alternative analysis methodologies. Constrained spherical harmonic deconvolution demonstrates potential, however current data-driven calibration (optimized for white matter) produce spurious peaks in the fibre orientation density, adversely affecting tractography. With clinical application in mind, we demonstrate an automated tissue-specific calibration which, for the first time, successfully reconstructs complex muscle tissue in vivo and include preliminary results of unsupervised tract segmentation.

Tim B. Dyrby¹, Henrik M Lundell¹, Matthew G Liptrot¹, Mark W Burke², Maurice Ptito^1,3, and Hartwig R Siebner^1
1Danish Research Centre for MR, Copenhagen University Hospital Hvidovre, Hvidovre, Denmark, ²College of Medicine, Howard University, Washington DC, United States, ³School of Optometry, University of Montreal, Montreal, Canada

We hypothesise that super resolution (SR) of HARDI DWI data can extract anatomical details that are usually obscured by partial volume effects and thus are currently only visible if acquired at higher spatial resolutions. Here we validate a simple SR approach, where DWI data are interpolated prior to reconstruction. Using 8x SR interpolation of low resolution images, acquired from an ex-vivo monkey brain, new anatomical details matching those in a high resolution dataset, became apparent. Higher SR-factors introduced smoothing. Our preliminary results suggest that HARDI DWI data contain hidden anatomical detail that can be extracted with this simple SR approach.

Marco Reisert¹, and Valerij Kiselev^1
1Medical Physics, University Medical Center Freiburg, Freiburg, Baden Württemberg, Germany

The accurate and reliable estimation of fiber orientation distributions based on diffusion-sensitized magnetic resonance images is a major prerequisite for tractography algorithms or any other derived statistical analysis. In this work we formulate the principle of fiber continuity (FC), which is based on the simple observation that the imaging of fibrous tissue implies certain expectations to the measured images. From this principle we derive a prior for the estimation of fiber orientation distributions based on high angular resolution diffusion imaging (HARDI). We demonstrate on simulated, phantom and in vivo data the superiority of the proposed approach.

Eric Aboussouan¹, Luca Marinelli², and Ek Tsoon Tan^2
1Barrow Neurological Institute, Phoenix, AZ, United States, ²GE Global Research, Niskayuna, NY, United States

Diffusion Spectrum Imaging (DSI) can be accelerated the application of compressed sensing (CS) in the three q-space dimensions. Rather than sampling a subset of a Cartesian q-space, we propose to further minimize coherent aliasing by undersampling q-space according to a non-Cartesian sampling pattern. Simulation and preliminary in vivo results are shown suggesting the validity of the approach.

Benoit Scherrer¹, and Simon K Warfield^1
1Radiology, Harvard Medical School, Boston, Massachusetts, United States

Multi-tensor models are of great interest for clinical applications because they enable the assessment of the white matter microstructure in addition to the brain connectivity. We propose a novel acquisition scheme and a novel fitting procedure for multi-fiber assessment. Our acquisition scheme combines spherical and cubic sampling. It enables multiple b-values to be acquired with low geometric and intensity distortion. Our optimization algorithm ensures spatially smooth and consistent positive definite tensors. We evaluate our CUSP-MFM (CUbe+SPhere Multi-Fiber Model) on both synthetic and clinical data. We demonstrate the ability of CUSP-MFM to characterize complex fiber structures from short duration acquisitions.

Greg Daniel Parker^1,2, and Derek K Jones^1
1CUBRIC, School of Psychology, Cardiff University, Cardiff, United Kingdom, ²School of Computer Science, Cardiff University, Cardiff, United Kingdom

Constrained spherical-harmonic deconvolution (CSD) provides resolution of intra-voxel “crossing fibre” orientation through deconvolution of an idealised (calibration) response from spherical harmonics fit to the target diffusion-weighted signal. Given a single-fibre low anisotropy target,we observe that, with high calibration anisotropies, spurious peaks are observed in the resultant deconvolution at elevations (relative to fibre orientation) consistent with the zero crossing (magic angle) and minima of the 2nd order Legendre polynomial, adversely affecting tractography. Bootstrap analysis however indicates that azimuthal distribution of such peaks appears uniform, allowing true “crossing fibre” to be distinguished through measurement of uncertainty.

Marta Morgado Correia¹, Guy B Williams², Frank Yeh³, Ian Nimmo-Smith¹, and Eleftherios Garyfallidis^1
1MRC Cognition and Brain Sciences Unit, Cambridge, United Kingdom, ²Wolfson Brain Imaging Centre, Cambridge, United Kingdom, ³Carnegie Mellon University, Pittsburgh, United States

Generalized Q-Sampling Imaging (GQI) has recently been introduced by Yeh and colleagues and was shown to have comparable accuracy to other well established q-space methods when it comes to resolving crossing fibres. In this study we compared the estimated values of MD, FA and Quantitative Anisotropy (QA) obtained with grid and shell GQI sampling schemes, in terms of their precision and ability to differentiate between different brain fibre populations. Our results suggest that a grid sampling scheme produces more robust results than a single shell acquisition.

Natasha Lepore*¹, Caroline Brun*², Maxime Descoteaux³, Yi-Yu Chou⁴, Greig de Zubicaray⁵, Katie McMahon⁵, Margie Wright⁶, Nicholas Martin⁶, James Gee², and Paul Thompson *equal contribution^7
1Department of Radiology, Children's Hospital, Los Angeles, Los Angeles, CA, United States, ²Department of Radiology, Penn Image Computing and Science Laboratory, University of Pennsylvania, Philadelphia, PA, United States, ³Université de Sherbrooke, Canada, ⁴Laboratory of NeuroImaging, UCLA, United States, ⁵University of Queensland, Australia, ⁶Genetic Epidemiology Lab, QIMR, Australia, ⁷Laboratory of NeuroImaging, UCLA, Los Angeles, CA, United States

Diffusion MRI is a popular tool used to compare brain white matter structure between groups of subjects. When this technique is used to compare two populations, it is common practice to reduce the sometimes-large number of diffusion gradients to a univariate measure at each voxel, such as the fractional anisotropy. However, we and others have designed voxel-wise comparison methods for HARDI data and used multivariate measures for HARDI group comparisons. Here we compare statistical power for two scalar and two multivariate measures derived from the HARDI signal.

Jian Cheng^1,2, Sylvain Merlet², Aurobrata Ghosh², Emmanuel Caruyer², Tianzi Jiang¹, and Rachid Deriche^2
1Institute of Automation, Chinese Academy of Sciences, Beijing, Beijing, China, People's Republic of, ²INRIA Sophia Antipolis, Sophia Antipolis, Sophia Antipolis, France

Since Diffusion Tensor Imaging (DTI) cannot detect the fiber crossing, many new works beyond DTI has been proposed to explore the q-space. Most works, known as single shell High Angular Resolution Imaging (sHARDI), focus on single shell sampling and reconstruct the Orientation Distribution Function (ODF). The ODF, which has no radial information at all, is just one of features of Ensemble Average Propagator (EAP). Diffusion Spectrum Imaging (DSI) is a standard method to estimate EAP via numerical Fourier Transform (FT), which needs lots of samples and is impractical for clinical study. Spherical Polar Fourier Imaging (SPFI) [1,2] was proposed to represent the signal using SPF basis, then the EAP and the ODF have analytical closed forms. So the estimation of the coefficients is very important. In [1,2], the coefficients are estimated based on a standard Least Square (LS) with L2 norm regularization (L2-L2). In this paper, we propose to estimate A using LS with L1 norm regularization (L2-L1), also named as Least Absolute Selection and Shrinkage Operator (LASSO). And we prove that the L2-L1 estimation of the coefficients is actually the well known Compressive Sensing (CS) method to estimate EAP, which brings lots of Mathematical tools and possibility to improve the sampling scheme in q-space.

Martin David King¹, Daniel C Alexander², David G Gadian¹, and Chris A Clark^1
1Institute of Child Health, University College London, London, United Kingdom, ²Computer Science, University College London, London, United Kingdom

Poor spatial resolution is a limitation in various diffusion MRI applications, including tractography. A Bayesian latent variables random effects model has been developed for increasing effective spatial resolution, based on a Markov random field treatment in which intrinsic Gaussian autoregressive priors are assigned to the fibre spherical coordinates. The model is used to separate crossing-fibres at the junction between the cingulum and corpus callosum, using diffusion MRI data acquired with a moderate b-value and 20 directions. The analyses were performed using Markov chain Monte Carlo simulation. Results demonstrate that a satisfactory separation of the crossing components can be obtained.

Jian Cheng^1,2, Aurobrata Ghosh¹, Tianzi Jiang², and Rachid Deriche^1
1INRIA Sophia Antipolis, Sophia Antipolis, Sophia Antipolis, France, ²Institute of Automation, Chinese Academy of Sciences, Beijing, Beijing, China, People's Republic of

In Diffusion Tensor Imaging (DTI), Riemannian framework (RF) has been proposed for processing tensors, which is based on Information Geometry theory. Recently RF also has been proposed for Orientation Distribution Function (ODF) computing. In this paper, we propose the RF for EAPs and implement it via SPFI. We proved that the RF for EAPs is diffeomorphism invariant, which is the natural extension of affine invariant RF for tensors. It could avoid the so-called swelling effect for interpolating EAPs, just like the RF for tensors. We also propose the Log-Euclidean framework (LEF), Affine-Euclidean framework (AEF), for fast processing EAPs, and Geometric Anisotropy (GA) for measuring the anisotropy of EAPs, which are all the extensions of previous concepts in RM for tensors respectively.

Ameer Pasha Hosseinbor¹, Moo K. Chung², Yu-Chien Wu³, and Andrew L. Alexander^4
1Medical Physics, University of Wisconsin-Madison, Madison, Wisconsin, United States, ²Biostatistics, University of Wisconsin-Madison, ³Radiology, University of Wisconsin-Madison, ⁴Medical Physics, University of Wisconsin-Madison

We present a novel technique for analytical EAP reconstruction from multiple q-shell acquisitions. The solution is based on the heat equation estimation of signal for each shell acquisition.

Anna Varentsova¹, Shengwei Zhang², and Konstantinos Arfanakis^2
1Biological, Chemical and Physical Sciences, Illinois Institute of Technology, Chicago, IL, United States, ²Biomedical Engineering, Illinois Institute of Technology, Chicago, IL, United States

The present work is devoted to the development of a high angular resolution diffusion imaging (HARDI) template from 67 coregistered, artifact-free datasets acquired with low angular resolution diffusion imaging. Preliminary results show that intravoxel fiber crossings can be resolved from combination of the 67 datasets, and that the information contained in the resulting template is in agreement with underlying fiber anatomy of the human brain.

Gemma L Morgan¹, Hui Zhang¹, Brandon Whitcher², and Daniel C Alexander^1
1Centre for Medical Image Computing, Department of Computer Science, University College London, London, United Kingdom, ²Clinical Imaging Centre, GlaxoSmithKline, London, United Kingdom

We present a new framework for modelling the regional variation of tissue parameters estimated from diffusion MRI across white matter regions of interest. The method can be used to compare differences in parameter variation between groups and statistical tests allow the localisation of significant differences. We demonstrate the technique on the midsagittal corpus callosum and are able to detect significant differences in the genu of two distinct age groups

Véronique Brion¹, Olivier Riff¹, Irina Kezele¹, Maxime Descoteaux², Denis Le Bihan¹, Jean-François Mangin¹, Cyril Poupon¹, and Fabrice Poupon^1
1NeuroSpin, CEA/I²BM, Gif-sur-Yvette, France, ²Sherbrooke University, Sherbrooke, Canada

We adressed the problem of the correction of the Rician noise, corrupting diffusion-weighted images at high b-values, in real-time. We combined a Linear Minimum Mean Square Error Estimator (LMMSE) together with a Kalman framework in order to compute in real-time the noise-free diffusion data, as well as the diffusion maps stemming from any local high angular resolution diffusion (HARDI) or hybrid diffusion (HYDI) model. A feedback is retropropagated from the Kalman filter to the LMMSE, in order both to reinforce the influence of the local structure onto the noise correction, and to prevent smoothing effects. The technique vas validated on synthetic and real data acquired at low signal to noise ratio (SNR) to assess its efficiency and the full pipeline was tested on the computation of orientation distribution functions.

Eizou Umezawa¹, Masayuki Yamada¹, Chiaki Tsunetomi¹, and Hirofumi Anno^1
1Graduate School of Health Sciences, Fujita Health University, Toyoake, Aichi, Japan

In diffusion MR analyses, the inversion symmetry of the water molecule diffusion is often assumed: the property of diffusion into a direction is identical to that into the opposite direction. Recently, a novel method, multi-shelled q-ball imaging (MS-QBI), has been proposed. In this study, we propose a method to detect the inversion asymmetry of diffusion with MS-QBI. We also perform the numerical simulation of detecting the asymmetry and examine the ability. MS-QBI will be able to detect the inversion asymmetry if the asymmetry is large enough. It is possible that MS-QBI can detect the inversion asymmetry in the case of the small voxel size realized by a high sensitivity coil system such as the cryoprobe.

Shani Ben Amitay¹, Silvia De Santis², Derek Jones², and Yaniv Assaf^3
1Tel Aviv University, Tel Aviv, Israel, ²CUBRIC, School of Psychology, Cardiff University, Wales, UK, United Kingdom, ³Tel Aviv University, Israel

High b-value diffusion imaging has been suggested to provide an enhanced contrast toward different cellular components. However, they suffer from low SNR, therefore, we suggest a framework based on both experimental data (DTI) and simulations (using CHARMED framework) to register the high b-value diffusion images. We inversely used the CHARMED model to generate template images that simulated the contrast seen in diffusion weighted images acquired at different b-values and gradient directions in the native image space of each subject. All high b-value images were registered to the matching templates. This approach makes motion correction feasible for the first time.

Nicholas Lange^1,2, and Peter J Basser^3
1Departments of Psychiatry and Biostatistics, Harvard University, Boston, MA, United States, ²Neurostatistics Laboratory, McLean Hospital, Belmont, MA, United States, ³PPITS, STBB, NICHD, National Institutes of Health, Bethesda, MD, United States

More biological information may already be contained in diffusion tensors measured typically by standard protocols run on clinical scanners in addition to that conveyed by their first-order size (e.g., mean, axial and radial diffusivity), second-order shape (e.g., FA) and third-order shape (skewness) coefficients. An enriched understanding of the “size” and “shape” of a tensor can be obtained by close inspection of its associated covariance tensor. This tensor takes on a variety of patterns whose complexity depends on the tissue’s unknown symmetry class, which can be determined by statistical methods, that may provide new insights in typical and disordered brain circuitry.

Jonathan D Clayden^1
1Institute of Child Health, University College London, London, United Kingdom

Here we demonstrate how probabilistic neighbourhood tractography (PNT) and principal component analysis (PCA) may be used together to analyse properties of white matter tracts in a robust and data-efficient manner. PNT is a method for segmenting tracts in groups using a shape model, while PCA allows common factors across tracts to be identified. This approach can help reduce the multiple comparisons problems widely faced in the statistical analysis of magnetic resonance data.

Cagatay Demiralp¹, and David H. Laidlaw^2
1Brown University, Providence, RI, United States, ²Brown University

We provide a proof of concept for modeling configuration distributions in DTI and their practical estimations. The power of the MAP-MRF framework comes from its mathematical convenience in modeling prior distributions and the fact that it results in a global optimization driven by local patches (context).

Yong Wang¹, and Sheng-Kwei Song^2
1Radiology, Washington University, Saint Louis, MO, United States, ²Radiology, Washington University in St. Louis, Saint Louis, MO, United States

Diffusion tensor imaging (DTI) derived radial diffusivity and axial diffusivity have been used to detect myelin and axon integrity respectively in CNS injury. Although it has been well recognized that crossing fibers pose significant challenge in the application of DTI derived directional diffusivity. It is not clear how the inflammation caused vasogenic edema and cell infiltration impact DTI findings. In this study, Monte Carlo simulation is employed to demonstrate the possible false assessment of axonal injury and/or demyelination of a normal axon fiber bundle under the influence of inflammation.

Derek K Jones^1
1CUBRIC, School of Psychology, Cardiff University, Cardiff, Wales, United Kingdom

We characterise the relative sensitivity of different white matter indices to the effects of partial volume contamination by CSF. We show that mean diffusivity is far more sensitivity than fractional anisotropy. The sensitivity is also dependent on the b-value and the anisotropy of the tensor.

Ivan I. Maximov¹, Farida Grinberg¹, and Nadim Jon Shah^1,2
1Institute of Neuroscience and Medicine 4, Forschungszentrum Juelich, Juelich, Germany, ²Department of Neurology, Faculty of Medicine, JARA, RWTH Aachen University, Aachen, Germany

We propose a new algorithm for diffusion tensor reconstruction based on the robust least median squares estimator. The developed method allows one to determine the outliers that may appear due to physiological noise or other tissue-related singularities. This feature and the stability of the least median squares to noise level variation yield a new and very promising approach in research and in clinical practice.

Firouzeh Tannazi¹, Lindsay Walker¹, Michael Curry¹, and Carlo Pierpaoli^1
1STBB/PPITS/NICHD/NIH, Bethesda, MD, United States

In this study we investigate the effects of the number of images comprising the DTI data set on the statistical properties of diffusion tensor eigenvalues and anisotropy indices derived from them. The results of Monte Carlo simulations along with the analysis of phantom DTI data indicate an overall underestimation of Eig3 and an overestimation of FA and Eig1 as the number of images in DTI estimation is reduced.

Shu-Wei Sun^1,2
1Biophysics and Bioengineering, Loma Linda University, Loma Linda, CA, United States, ²Radiation Medicine, Loma Linda University, Loma Linda, CA, United States

Diffusion Tensor Imaging (DTI) is achieved by collecting a series of Diffusion Weighted Images (DWI) with different diffusion weighted vectors. For each DWI, it is usually acquired with multiple repetitions to boost the signal to noise ratio. On may perform k-space average or image-space average to boost the signal to noise ratio for each DWI. In this study, we compared DTI maps of mouse brains in vivo using k-space average, image-space average, and no average approaches. K-space average provided the least contrast for presenting white matter on RA maps.

Jens H Jensen^1
1Department of Radiology, New York University School of Medicine, New York, NY, United States

The standard theories for vessel size imaging (VSI) and microvessel density imaging (MDI) are based on an assumption of isotropic water diffusion. Therefore, the validity of applying these techniques to white matter may be questioned. Here the corrections to the basic VSI and MDI equations due to diffusion anisotropy are explicitly calculated in terms of the diffusion tensor. For the most anisotropic brain regions, the corrections are found to be significant, although not large. These results may relevant for application of VSI and MDI to the assessment of angiogenesis in white tumors.

Siawoosh Mohammadi¹, Zoltan Nagy¹, Harald E Moeller², David Carmichael^3,4, Mark Symms³, Oliver Josephs¹, and Nikolaus Weiskopf^1
1Wellcome Trust Centre for Neuroimaging, UCL Institute of Neurology, University College London, London, United Kingdom, ²Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, ³Clinical and Experimental Epilepsy, UCL Institute of Neurology, London, United Kingdom, ⁴MRI unit, National Society for Epilepsy, Chalfont St. Peter, United Kingdom

Imaging artefacts, which perturb diffusion-weighted images, can bias the estimated diffusion tensor. Important sources of imaging artefacts in DTI are eddy current fields, gradient nonlinearities or mis-calibration of the gradient amplitude. They can be modelled by introducing the concept of a local perturbation field (LPF). In this study, we used first-order perturbation theory to introduce a physically-informed model that estimates and corrects for the effect of LPFs. Using phantom and human DTI measurements on two different scanners we were able to estimate the LPFs and improve the quality of FA maps without requiring vendor-specific information.

Christopher L Welsh^1,2, Edward W Hsu^1,2, and Edward VR DiBella^1,2
1Bioengineering, University of Utah, Salt Lake City, Utah, United States, ²UCAIR, University of Utah, Salt Lake City, Utah, United States

Diffusion Tensor Imaging (DTI) is useful for characterizing tissue microstructure, but suffers from long scan time and low SNR. To allow faster acquisition, a model-based strategy is presented to directly estimate diffusion tensors from undersampled k-space data. Using an acceleration factor of 2, different sampling schemes were investigated and found to generally outperform acquiring equivalent number of full-resolution scans. Minor performance differences were also observed among the schemes for estimating different DTI parameters. These findings suggest the proposed strategy can be used to reduce DTI scan time by half while incurring little or no loss in the parameter estimation accuracy.

Ben Jeurissen¹, Maarten Naeyaert¹, Alexander Leemans², and Jan Sijbers^1
1Vision Lab, Dept. of Physics, University of Antwerp, Antwerp, Belgium, ²Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands

Misalignment between diffusion weighted (DW) images and the corresponding gradient orientations leads to misinterpretation of rotationally variant diffusion parameters and tractography results. We propose a method to automatically correct for rigid misalignment of the gradient orientations, using a registration metric based on the average fiber trajectory length. Using simulations we show that our method converges to the 'ground truth' orientations and that small 'angulation' errors can still be detected, that are easily overlooked by visual inspection. While the method uses diffusion tensor tractography to calculate trajectory lengths, the recovered gradient orientations can be applied in general DWI post-processing.

Sjoerd B Vos¹, Max A Viergever¹, and Alexander Leemans^1
1Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands

In order to reduce DTI scan time, slice thickness is often increased while maintaining a high in-plane resolution. The resulting anisotropic voxel size means that the orientation-dependent diffusion information is sampled in a higher resolution in-plane than through-plane. In this work we show that the bias in fractional anisotropy, introduced by anisotropically acquired DTI data, is dependent on the orientation of the fiber bundle of interest. Furthermore, we demonstrate that the diffusion bias is not solely determined by the anisotropy of the voxel size, but strongly depends on the relation between the fiber bundle and the data grid as well.

KI SUENG CHOI^1,2, Alexandre R. Franco², Paul E. Holtzheimer², Helen S. Mayberg², and Xiaoping P. Hu^1
1Bioengineering, Georgia Institute of Technology / Emory University, Atlanta, GA, United States, ²Psychiatry and Behavioral Sciences, Emory University, Atlanta, GA, United States

DW-EPI is very sensitive to B0 inhomogeneities that produce geometric distortion, primarily along the phase-encoding direction. As a result, artifacts degrade the ability of using diffusion measures and diffusion tractography. Several different techniques were suggested for correcting susceptibility distortion. However, to apply these correction techniques, additional images (field map, T2) are required. In this work, we examine susceptibility distortion correction using image based non-linear registration along with inverting the intensity of the T1 image. This method exhibits better performance for geometrical distortion correction in frontal region and an ability to analyze clinical diffusion data without additional collection of images.

Wim Van Hecke^1,2, Louise Emsell^3,4, Alexander Leemans⁵, Caroline Sage⁶, Jelle Veraart⁷, Stefan Sunaert⁶, Jan Sijbers⁷, and Paul M Parizel^7
1University of Antwerp, Antwerp, Antwerp, Belgium, ²University of Leuven, Leuven, Leuven, Belgium, ³The Murdoch Childrens Research Institute, Australia, ⁴NUI Galway, Ireland, ⁵Image Sciences Institute, Utrecht, Netherlands, ⁶University of Leuven, Belgium, ⁷University of Antwerp, Belgium

In this study, the effect of atlas selection on voxel based analyses of DTI data is examined. To this end, data sets of healthy subjects and multiple sclerosis patients were aligned to a population-specific atlas, a subject-based MNI template, and the ICBM-81 atlas.

Kimihiro Ogisu¹, Hidetsugu Sakai², and Toru Yamamoto^2
1Graduate School of Medicine, Hokkaido University, Sapporo, Japan, ²Graduate School of Health Sciences, Hokkaido University

The signal from intravoxel incoherent motion (IVIM) appears in diffusion-weighted imaging at low b values. This low-b-value component is believed to reflect capillary blood flow in tissue. However, the quantity of the low-b-value component contradicts the data on capillary blood volume in the literature. To determine the origin of the low-b-value component, we investigated the proton density and T2 value of the component. We found that the low-b-value component has a larger T2 value than does blood and that this component exhibits IVIM. Because the hydrostatic pressure and osmotic pressure in capillaries drive interstitial fluid flow, which has a large T2 value, we suggest that the interstitial fluid mainly contributes to the low-b-value component.

Saurav Chandra¹, Sergio Caballero², Maria B Grant², and John R Forder^1,3
1Biomedical Engineering, University of Florida, Gainesville, FL, United States, ²Pharmacology, University of Florida, ³Radiology, University of Florida

The study aims to restore visual function in mice impaired by retinal degeneration by using hematopoietic stem cells (HSCs) to repair the retinal pigment epithelium (RPE) layer. HSCs expressing the gene RPE65 were successful in differentiating into RPE cells and proliferating to the specific site to restore the RPE layer. Diffusion tensor imaging successfully visualized differences between repaired and unrepaired RPE layers in mice. Fractional anisotropy (FA) was observed to markedly higher in repaired retinas (as is the case in any structured anatomy) compared to unrepaired retinas which had a depleted RPE layer.

Saurav Chandra¹, Angelos Barmpoutis², Nicholas Simpson³, and John R Forder^1,4
1Biomedical Engineering, University of Florida, Gainesville, FL, United States, ²Computer and Information Sciences Engineering, University of Florida, ³College of Medicine, University of Florida, Gainesville, FL, United States, ⁴Radiology, University of Florida, Gainesville, FL, United States

Diabetic retinopathy is the most common eye disease affecting diabetics and a leading cause of blindness. However, it cannot be diagnosed in its early stages. We used High Angular Resolution Diffusion Microscopy (HARDM) as a non-invasive tool to detect this disease at an early stage in mice. HARDM of control eyes showed water diffusion in the retina was restricted, reflecting an organized structure within the retinal layers. Comparison with control eyes showed the integrity of these layers is compromised in eyes from diabetic animals with elevated glucose levels. FA is also significantly decreased in the diabetic retinas compared to controls.

Mustafa Okan Irfanoglu¹, Lindsay Walker², Raghu Machiraju³, and Carlo Pierpaoli^2
1Computer Sciences and Engineering, The Ohio State University, Columbus, OH, United States, ²NIH, ³The Ohio State University

In this work, we propose a novel technique to account for changes in signal variance in diffusion weighted images due to interpolation artifacts related to image registration based motion and distortion correction steps. Our technique can model signal variance changes due to sequential transformation applications and can cope with all types of interpolation kernels. We show our results as improvements in Chi-squared maps of tensor fitting.