Sharpening Improves Clinically Feasible Q-Ball Imaging Reconstructions

Slides:



Advertisements
Similar presentations
Probabilistic Inverse Dynamics for Nonlinear Blood Pattern Reconstruction Benjamin Cecchetto, Wolfgang Heidrich University of British Columbia.
Advertisements

Medical Image Reconstruction Topic 4: Motion Artifacts
In Chan Song, Ph.D. Seoul National University Hospital
NON-EXPONENTIAL T 2 * DECAY IN WHITE MATTER P. van Gelderen 1, J. A. de Zwart 1, J. Lee 1,3, P. Sati 1, D. S. Reich 1, and J. H. Duyn 1. 1 Advanced MRI.
2D/3D Shape Manipulation, 3D Printing
Camino and DTI-TK: Advanced Diffusion MRI Pipeline for Traumatic Brain Injury Gary Hui Zhang, PhD Microstructure Imaging Group Centre for Medical Image.
DTI group (Pitt) Instructor: Kevin Chan Kaitlyn Litcofsky & Toshiki Tazoe 7/12/2012.
DIFFUSION TENSOR IMAGING
Lighting affects appearance. What is the Question ? (based on work of Basri and Jacobs, ICCV 2001) Given an object described by its normal at each.
MaxEnt 2007 Saratoga Springs, NY
ADC and ODF estimation from HARDI
High Angular Diffusion Imaging and its Visualization… Limitations of DTI Why HARDI is better?!? Different HARDI models My ideas and current work.
Diffusion Tensor MRI And Fiber Tacking Presented By: Eng. Inas Yassine.
Diffusion Tensor Imaging Tim Hughes & Emilie Muelly 1.
Reproducibility of diffusion tractography E Heiervang 1,2, TEJ Behrens 1, CEM Mackay 3, MD Robson 3, H Johansen-Berg 1 1 Centre for Functional MRI of the.
Multiple testing Justin Chumbley Laboratory for Social and Neural Systems Research Institute for Empirical Research in Economics University of Zurich With.
Master thesis by H.C Achterberg
Introduction to diffusion MRI
Multiple testing Justin Chumbley Laboratory for Social and Neural Systems Research Institute for Empirical Research in Economics University of Zurich With.
Diffusion Physics H 2 O in the body is always in random motion due to thermal agitation B.M. Ellingson, Ph.D., Dept. of Radiological Sciences, David Geffen.
05/19/11Why’n’how | Diffusion model fitting and tractography0/25 Diffusion model fitting and tractography: A primer Anastasia Yendiki HMS/MGH/MIT Athinoula.
J OHNS H OPKINS U NIVERSITY S CHOOL O F M EDICINE Statistically-Based Reorientation of Diffusion Tensor Field XU, D ONGRONG S USUMU M ORI D INGGANG S HEN.
Introduction to Power Spectrum Estimation Lloyd Knox (UC Davis) CCAPP, 23 June 2010.
13.1 Fourier transforms: Chapter 13 Integral transforms.
Rician Noise Removal in Diffusion Tensor MRI
TSTAT_THRESHOLD (~1 secs execution) Calculates P=0.05 (corrected) threshold t for the T statistic using the minimum given by a Bonferroni correction and.
Technical Factors affecting Apparent Diffusion Coefficient in Women with Locally Advanced Cervical Cancer –Goal: To measure ADC variability resulting from.
Brain Lab Imaging Didactics
1/25 Current results and future scenarios for gravitational wave’s stochastic background G. Cella – INFN sez. Pisa.
Benoit Scherrer, ISBI 2010, Rotterdam Why multiple b-values are required for multi-tensor models. Evaluation with a constrained log- Euclidean model. Benoit.
Trade-offs between Angular and Spatial Resolution in High Angular Resolution Diffusion Imaging Liang Zhan 1, Neda Jahanshad 1, Alex D. Leow 2,3, Matt A.
Tensor Distribution Function in Multiple Shell High Angular Resolution Diffusion Imaging Tensor Distribution Function in Multiple Shell High Angular Resolution.
Processing HARDI Data to Recover Crossing Fibers Maxime Descoteaux PhD student Advisor: Rachid Deriche Odyssée Laboratory, INRIA/ENPC/ENS, INRIA Sophia-Antipolis,
Benoit Scherrer, ISBI 2011, Chicago Toward an accurate multi-fiber assessment strategy for clinical practice. Benoit Scherrer, Simon K. Warfield.
HighlY Constrained Back PRojection (HYPR) Thank you to Oliver Wieben!!
Jordan Hamm (BA, BSc) University of Georgia, Athens, Georgia Alexandra Reichenbach (MSc, Dipl-Ing) Max Planck Institute for Biological Cybernetics, Tuebingen,
Instructor Kwan-Jin Jung, Ph.D. (Carnegie Mellon University) Technical Assistant Nidhi Kohli (Carnegie Mellon University) David Schaeffer (University of.
Lighting affects appearance. How do we represent light? (1) Ideal distant point source: - No cast shadows - Light distant - Three parameters - Example:
NA-MIC National Alliance for Medical Image Computing NAMIC UNC Site Update Site PI: Martin Styner Site NAMIC folks: Clement Vachet, Gwendoline.
Luke Bloy1, Ragini Verma2 The Section of Biomedical Image Analysis
Analytic ODF Reconstruction and Validation in Q-Ball Imaging Maxime Descoteaux 1 Work done with E. Angelino 2, S. Fitzgibbons 2, R. Deriche 1 1. Projet.
Fiber Demixing with the Tensor Distribution Function avoids errors in Fractional Anisotropy maps Liang Zhan 1,Alex D. Leow 2,3, Neda Jahanshad 1, Arthur.
Dave Frank & Maggie Mahan
NA-MIC National Alliance for Medical Image Computing Mathematical and physical foundations of DTI Anastasia Yendiki, Ph.D. Massachusetts.
NA-MIC National Alliance for Medical Image Computing NAMIC UNC Site Update Site PI: Martin Styner UNC Site NAMIC folks: C Vachet, G Roger,
Integrity of white matter in the corpus callosum correlates with bimanual co-ordination skill Heidi Johansen-Berg 1, Valeria Della-Maggiore 3, Steve Smith.
References [1] Coupé et al., An optimized blockwise nonlocal means denoising filter for 3-D magnetic resonance images. IEEE TMI, 27(4):425–441, 2008.
Lighting affects appearance. Lightness Digression from boundary detection Vision is about recovery of properties of scenes: lightness is about recovering.
Measuring Water Diffusion In Biological Systems Using Nuclear Magnetic Resonance Karl Helmer HST 583, 2006
Robust Regularization for the Estimation of Intra-Voxel Axon Fiber Orientations Also presented at MMBIA Anchorage, Jun, 2008 Alonso Ramirez-Manzanares.
Stochastic Background Data Analysis Giancarlo Cella I.N.F.N. Pisa first ENTApP - GWA joint meeting Paris, January 23rd and 24th, 2006 Institute d'Astrophysique.
DTI Acquisition Guide Donald Brien February 2016.
COMPARATIVE LATERALIZING ABILITY of MULTIMODALITY MR IMAGING in TEMPORAL LOBE EPILEPSY ¹ Karabekir Ercan, M.D. ¹ ¹ H.Pinar Gunbey, M.D. ¹ ¹ Elcin Zan,
Data analysis steps Pre-process images to reduce distortions
Diffusion Tensor MRI From Deterministic to Probabilistic Modelling
New Features Added to Our DTI Package XU, Dongrong Ph.D. Columbia University New York State Psychiatric Institute Support: 1R03EB A1 June 18, 2009.
Mapping White Matter Connections in the Brain using Diffusion-Weighted Imaging and Tractography Andy Alexander Waisman Center Departments of Medical Physics.
Super-resolution MRI Using Finite Rate of Innovation Curves Greg Ongie*, Mathews Jacob Computational Biomedical Imaging Group (CBIG) University of Iowa.
Non-linear Realignment Using Minimum Deformation Averaging for Single-subject fMRI at Ultra-high Field Saskia Bollmann1, Steffen Bollmann1, Alexander.
Degradation/Restoration Model
Introduction to diffusion MRI
Chapter 13 Integral transforms
Spatially Varying Frequency Compounding of Ultrasound Images
Introduction to diffusion MRI
Outline Linear Shift-invariant system Linear filters
Outline Linear Shift-invariant system Linear filters
Anisotropy Induced by Macroscopic Boundaries: Surface-Normal Mapping using Diffusion-Weighted Imaging  Evren Özarslan, Uri Nevo, Peter J. Basser  Biophysical.
Diffusion MRI of Complex Neural Architecture
Images reconstruction. Radon transform.
Presentation transcript:

Sharpening Improves Clinically Feasible Q-Ball Imaging Reconstructions Maxime Descoteaux & Rachid Deriche Project Team Odyssee INRIA Sophia Antipolis, France Presente-toi Parle des groupes

Improving angular resolution of Q-Ball Imaging Can we transform the diffusion ODF (dODF) into a sharp fiber ODF (fODF)? dODF dODF min-max fODF dODF min-max

In the literature… Fiber orientation density (FOD) function Spherical Deconvolution = HARDI Signal Fiber response function FOD [Tournier et al 2004-2005-2006-2007, Alexander et al 2005, Anderson 2005, Dell’Acqua et al 2007]

Sketch of the method = Convolution assumption

Sketch of the method A deconvolution approach FRT HARDI Signal dODF sharpening fODF

Step 1: Analytical ODF estimation [Descoteaux et al MRM 2006 & MRM 2007 accepted] Laplace-Beltrami regularized estimation of the HARDI signal [Anderson MRM 05, Hess et al MRM 06, Descoteaux et al RR 05, ISBI 06]

Step 2: Diffusion ODF kernel for deconvolution Estimate from real data Take 300 voxels with highest FA Assumed to contain a single fiber population Find average prolate tensor D that fits the data Diffusion ODF kernel is Analytical ODF [Tuch MRM 2004 Descoteaux RR 2005] where e1 > e2 are e-values of D and t := cos

Step 3: Deconvolution with the Funk-Hecke theorem Final sharp fiber ODF Linear transformation of the spherical harmonic coefficients describing the signal [Descoteaux et al Research Report 2005, MRM 2007 accepted.]

Summary of the method cj 2 Plj(0) fj Analytical FRT HARDI Signal dODF Deconvolution Sharpening fODF Analytical FRT cj 2 Plj(0) fj

Separation angle Sharp fiber ODF Min-max normalized ODF (Two-tensor model, FA1 = FA2 = 0.7, SNR 30, b-value 3000 s/mm2, 60 DWI)

Simulation results Sharpening improves angular resolution and improves fiber detection with small angular error on the detected maxima Mean angular error 4.5 +- 1.23 ~20 improvement

Real data acquisition N = 60 directions 72 slices, 128 x 128 1.7 mm3 voxels b-value 1000 s/mm2 Sharp fiber ODF estimation of order 4 in less than 20 seconds 3 Tesla Magnetom Trio scanner (Siemens, Erlangen) equipped with an 8-channel head array coil [47]. The spin-echo EPI sequence, TE = 100 ms, TR = 12 s, 128 x 128 image matrix, FOV = 220 x 220 mm2 , consists of 60 diffusion encoding gradients [56] with a b-value of 1000 s/mm2 and 7 images without any diffusion weightings, which are evenly distributed. The measurement of 72 slices with 1.7mm thickness (no gap), which covered the whole brain, was repeated three times, resulting in an acquisition time of about 45 minutes. Hence, the S0 image is the average of 21 b = 0 images. The SNR in the white matter of the averaged b = 0 image S0 was estimated to approximately 37. Additionally, fat saturation was employed, 6/8 partial Fourier imaging, Hanning window filtering, and parallel GRAPPA imaging with a reduction factor of 2. [Thanks to Max Planck Institute, Leipzig, Germany]

Crossing voxel between motor stripe and SLF Unequal volume fraction of the 2 fiber compartments Voxel manually chosen by expert.

Real data - Crossing between the cc, cst, slf fODFs dODFs diffusion tensors

Take home message It is possible to transform the diffusion ODF into a sharp fiber ODF for clinical QBI acquisitions Method is: Linear, fast, analytic, robust to noise All this possible because of the properties of the spherical harmonics and the Funk-Heck theorem

Current work and perspectives… Compare with spherical deconvolution Study the link between the two approaches Study the negative lobe problem that appears with spherical deconvolution [see Tournier et al 2007, Sakaie et al 2007 and Dell’Acqua et al 2007] Use the fiber ODF for tracking Deterministic Probabilistic

Thank You! Key References: Thanks to: Descoteaux et al, Regularized, Fast and Robust Analytical Q-Ball Imaging, MRM 2007 Descoteaux et al, ISBI 2006 & INRIA Research Report 2005 D. Tuch, Q-Ball Imaging, MRM 2004 Tournier et al, … Spherical Deconvolution…, NeuroImage 2004 & 2007 http://www-sop.inria.fr/odyssee Thanks to: -A. Anwander & T. Knosche of the Max Planck Institute, Leipzig, Germany -C. Poupon et al, Neurospin, Saclay, Paris

Why is this important? Better fiber detection with QBI Possible to recover crossings on datasets acquired in clinical setup Better tracking Even when b-value is small and we know QBI approximation

Step 1: Analytical ODF estimation [Tuch MRM 04] HARDI Signal FRT -> dODF Funk-Radon Transform (FRT)

SNR effect Sharp fiber ODF Min-max normalized ODF (Two-tensor model, FA1 = FA2 = 0.7, seperation angle 60, b-value 3000 s/mm2, 60 DWI)

Volume fraction effect Sharp fiber ODF Min-max normalized ODF (Two-tensor model, FA1 = FA2 = 0.7, seperation angle 60, SNR 30, b-value 3000 s/mm2, 60 DWI)