Diffusion Controversies

Thursday 15 May 2014    13:30 - 14:30

Space 1/Power Poster Theatre & Traditional Poster Hall
Moderators: Joseph J. H. Ackerman, Ph.D. & Peter J. Basser, Ph.D., A.B., S.M.

Andreas Wetscherek¹ and Frederik Bernd Laun^1,2
1Medical Physics in Radiology (E020), DKFZ, Heidelberg, Germany, ²Quantitative Imaging-Based Disease Characterization (E011), DKFZ, Heidelberg, Germany

The intravoxel incoherent motion signal calculated from normalized phase distributions is compared to the signal obtained using Gaussian phase approximation (GPA). For physiological parameters found in liver and pancreas, the GPA breaks down for b-values > 50 s/mm² for flow compensated gradient profiles or when a parabolic distribution of blood flow velocities is assumed. Moreover It is shown that the pseudo-diffusion coefficient D* as defined by Le Bihan is typically underestimated when calculated from a biexponential fit. Since the GPA doesn’t hold in general, the use of normalized phase distributions is strongly recommended for quantification of intravoxel incoherent motion parameters.

Susie Y. Huang¹, Aapo Nummenmaa¹, Thomas Witzel¹, Tanguy Duval², Julien Cohen-Adad², Lawrence L. Wald^1,3, and Jennifer A. McNab^4
1Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, Harvard Medical School, Boston, MA, United States, ²Institute of Biomedical Engineering, Ecole Polytechnique de Montreal, Montreal, QC, Canada, ³Harvard-MIT Division of Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, United States, ⁴Richard M. Lucas Center for Imaging, Department of Radiology, Stanford University, Stanford, CA, United States

Translating diffusion MRI methods for axon diameter mapping to clinical applications requires higher maximum gradient strengths (Gmax) than are currently available on commercial scanners. Using a dedicated high-gradient 3T MRI scanner with Gmax=300mT/m, we systematically study the effect of gradient strength on in vivo axon diameter estimates in the human corpus callosum. We find that an optimal q-space sampling scheme should incorporate the highest possible gradient strengths and draw from a wide range of gradient strengths and diffusion times. The improvement in axon diameter estimates will inform protocol development and encourage the adoption of higher gradient systems for widespread use.

Gonzalo Sanguinetti¹ and Rachid Deriche^1
1Athena Project-Team, INRIA, Sophia-Antipolis, PACA, France

We propose an original technique for measuring axon diameter distributions that simplifies the AxCaliber framework in complexity and shortens the acquisition time. In particular, we derive a new closed form expression describing the echo attenuation in the intra-axonal space. The method is validated using Monte-Carlo simulations of NMR signals from complex white matter-like environments. The results are promising and illustrate the potential of the method.

Eric Aboussouan¹, Rafael O'Halloran¹, Murat Aksoy¹, Daniel kopeinigg¹, and Roland Bammer^1
1Radiology, Stanford University, Stanford, CA, United States

High-resolution diffusion-weighted imaging is limited to multi-shot acquisitions, which are problematic due to inter-shot phase variations caused by rigid-body (non-repeatable) and pulsatile (repeatable over the cardiac cycle) motion during the diffusion-encoding periods. The purpose of this work is to measure the 3D non-linear component of the phase in a novel pulsatile phantom with repeatable motion and prospectively correct the linear and non-linear components of this spurious phase using a 3D RF pulse.

Hing-Chiu Chang¹, Shayan Guhaniyogi¹, and Nan-Kuei Chen^1
1Brain Imaging and Analysis Center, Duke University Medical Center, Durham, North Carolina, United States

Progresses in MRI based connectivity network mapping for translational neuroimaging is currently limited by the spatial resolution that can be achieved with conventional DTI protocols. The recent progress in 3D multi-slab EPI sequence makes it possible to acquire human DTI data with 1.3mm3 isotropic voxel size. However, the in vivo human brain DTI at sub-millimeter isotropic resolution, to our knowledge, has not yet been routinely achieved yet. We integrated the 3D multi-slab EPI acquisition and the multiplexed sensitivity encoding (MUSE) post-processing algorithm, to acquire high-quality and high-SNR human brain DTI data in vivo at high spatial resolution: 0.85mm^3.

Bibek Dhital¹, Jochen Leupold², and Valerij G. Kiselev^2
1German Cancer Consortium (DKTK), Heidelberg, Baden, Germany, ²Department of Diagnostic Radiology, University Medical Center Freiburg, Freiburg, Baden Württemberg, Germany

In this abstract we discuss application of PFG at long diffusion times and small q to discriminate between and hindered and restricted diffusion pools. We implement this as a preparation module to oscillating gradient diffusion weighted sequence. Since both the PFG and oscillating gradient waveforms are sensitive to diffusive motion, proper application of such ‘PFG-filter’ requires that no cross terms exist between the two. Two sets of oscillating gradient diffusion weighted measurements on a celery stalk, one with the PFG-filter and other without. Our results confirm that in voxels containing heterogeneous regions, the PFG-filter can resolve the two compartments.

Evren Ozarslan¹, Alexandru V Avram², Peter J Basser², and Carl-Fredrik Westin^1
1Department of Radiology, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, United States, ²Section on Tissue Biophysics and Biomimetics, PPITS, NICHD, National Institutes of Health, Bethesda, MD, United States

Rotating field gradients (RFGs), generated by applying sine- and cosine-modulated gradient waveforms along two perpendicular directions, provide an alternative diffusion sensitization mechanism for MRI. Two RFGs with a 90 degree phase shift, applied around the 180 degree RF pulse in a spin echo sequence can be used to measure the dODF directly, obviating the need to transform the data from a space reciprocal to the displacement space. Experiments on an asparagus specimen performed on a clinical scanner confirmed the predicted signal dependence in anisotropic environments.

Noam Shemesh¹, Gonzalo A Alvarez¹, and Lucio Frydman^1
1Chemical Physics, Faculty of Chemistry, Weizmann Institute of Science, Rehovot, Israel

Noninvasively characterizing cellular size distributions is paramount since their features impact underlying functional or biological capacities. Herein, we harness Non-uniform-Oscillating-Gradient Spin-Echo Magnetic-Resonance-Imaging (NOGSE MRI) – a methodology probing diffusion dynamics with extraordinary sensitivity towards compartmental dimensions – for depicting size distributions in a simple, one-dimensional experiment. Simulations and experiments in yeast cells validate NOGSE’s ability to faithfully reconstruct cellular size distributions; NOGSE-derived maps of size distribution properties in the mouse brain reveal hallmark microstructural features in both white and gray matter. NOGSE’s exquisite sensitivity towards length to the power of six renders it highly promising for contrasting disease in-vivo from size distribution contrast.

David Andrew Porter^1
1Healthcare Sector, Siemens AG, Erlangen, Germany

Readout-segmented EPI (rs-EPI) is an alternative sequence for diffusion-weighted imaging of the brain with less artifact and higher resolution than single-shot EPI. The higher-quality, low-b-value image could in some cases be used to replace a separate T2-weighted acquisition and reduce overall examination time, but the echo time is usually shorter than that used in typical clinical protocols. This study introduces a modification to the sequence, which generates additional T2-weighted images with a user-specified echo time that is suitable for clinical studies. The technique can also provide integrated T2*-weighted images, which may be useful for detecting hemorrhage in acute stroke.

Samuel St-Jean¹, Pierrick Coupé², and Maxime Descoteaux^1
1Sherbrooke Connectivity Imaging Lab (SCIL), Université de Sherbrooke, Sherbrooke, Québec, Canada, ²Unité Mixte de Recherche CNRS (UMR 5800), Laboratoire Bordelais de Recherche en Informatique, Bordeaux, France

Diffusion Weighted Images datasets suffer from low SNR, especially at high b-values. High noise levels bias the measurements because of the non-Gaussian nature of the noise, which in turn can lead to a false and biased estimation of the diffusion parameters. We propose to use the redundancy of DWIs as a sparse representation to reduce the noise level and achieve a higher SNR using dictionary learning and sparse coding, without the need for additional acquisition time. We show quantitative results and compare with current state-of-the-art methods using perceptual metrics, diffusion metrics and ODFs reconstruction.

Zoltan Nagy^1,2, Tara Ganepola³, Martin I Sereno³, Nikolaus Weiskopf¹, and Daniel C Alexander^4
1Wellcome Trust Centre for Neuroimaging, University College London, London, United Kingdom, ²Laboratory for Social and Neural Systems Research, University of Zürich, Zürich, Switzerland, ³Department of Cognitive, Perceptual and Brain Sciences, University College London, United Kingdom, ⁴Center for Medical Image Computing, University College London, United Kingdom

MRI based in–vivo histology of brain tissue is an active research area with several approaches using different contrasts. Previously, we have used high angular resolution diffusion imaging data with a single b–value to construct a feature vector, which we proposed as a method for grey matter cortical parcellation. Here, we investigate the utility of combining data from several b–values (i.e. constructing a 2D feature matrix). The results strongly suggest that the combining information contained in these different datasets improves the parcellation. Future work will refine the choice of b–values and focus on histilogical validation.

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

Noisy MRI magnitude and root sum-of-squares (SoS) images follow the Rician and non-central chi distribution respectively. In diffusion MRI, estimation of diffusion parameters can be inaccurate due to the noise bias introduced by these distributions. This work presents a new approach to model and estimate HARDI parameters from Rician and non-central chi data. We show estimation results from both simulated and noisy real data, and demonstrate how this method can improve estimation compared to existing approaches.

Michael Paquette¹, Sylvain Merlet², Rachid Deriche², and Maxime Descoteaux^1
1Universite de Sherbrooke, Sherbrooke, Quebec, Canada, ²INRIA Sophia Antipolis, Sophia Antipolis, France

Diffusion Spectrum Imaging (DSI) is a well established method to recover the diffusion propagator (EAP). The orientation distribution function (ODF) is computed from this discretized EAP and used for tractography. However, there are several important implementation considerations that are tossed aside in the literature and the publicly available softwares. We investigate all the real steps necessary to go from the DSI signal to the ODF and provide applicable recommendations that greatly improve the accuracy of the local orientation detected. These recommendations come ”free-of-charge” as they are applicable to all existing DSI data and do not require a significant increase in computation time.

Van Jay Wedeen¹, Farzad Mortazavi², Ruopeng Wang¹, Wen-Yih Isaac Tseng³, Thomas Witzel¹, Aapo Nummenmaa¹, William Morrison⁴, H Eugene Stanley⁴, Lawrence Wald¹, and Douglas Rosene^2
1Radiology, Massachusetts General Hospital, Charlestown, MA, United States, ²Anatomy and Neurobiology, Boston University, Boston, MA, United States, ³Center for Optoelectronic Biomedicine, National Taiwan University, Taipei, Taiwan, ⁴Boston Unviersity, Boston, United States

Recently it has been found that the fiber pathways of the brain follow a curvilinear grid derived from the three axes of development. Here we present a method to objectively identify this structure among the brain pathways. For each voxel and pair of diffusion ODF maximum vectors, arrays of parallel crossing paths are constructed, and retained only where they form a 2D surface. In rhesus DSI, grid structure was demonstrated explicitly and extensively. In gyri, the preponderance of pathways followed rectilinear grid trajectories. These studies show grid structure can be objectively defined, and promise new quantitative strategies for MRI tractography.

Cibu P Thomas^1,2, Frank Q Ye³, Mustafa Okan Irfanoglu^1,2, Pooja D Modi¹, Kadharbatcha S Saleem³, David A Leopold³, and Carlo Pierpaoli^1
1National Institute of Child Health and Human Development, Bethesda, Maryland, United States, ²Center for Neuroscience and Regenerative Medicine, USUHS, Bethesda, Maryland, United States, ³National Institute of Mental Health, Bethesda, Maryland, United States

Although tractography based on diffusion-weighted magnetic resonance imaging (DWI) is a widely used technique for mapping the structural connections of the brain in vivo, its anatomical accuracy is highly questionable. It is generally assumed that this limitation can be resolved if DWI data with sufficiently high spatial and angular resolution and superior signal to noise ratio (SNR) can be acquired. Here, we report that despite using DWI data of unprecedented quality none of the tractography techniques we tested consistently showed superior anatomical accuracy. These results suggest that the quest for an optimal diffusion tractography technique may be elusive.