Méthodes d imagerie pour les écoulements et le CND

Similar documents
CIVA Computed Tomography Modeling

Adapted acquisition trajectory and iterative reconstruction for few-views CT inspection

Phase problem and the Radon transform

Limited view X-ray CT for dimensional analysis

GPU implementation for rapid iterative image reconstruction algorithm

MEDICAL IMAGE ANALYSIS

Acknowledgments and financial disclosure

Characterization of microshells experimented on Laser Megajoule using X-Ray tomography

Advanced Image Reconstruction Methods for Photoacoustic Tomography

Continuous and Discrete Image Reconstruction

3/27/2012 WHY SPECT / CT? SPECT / CT Basic Principles. Advantages of SPECT. Advantages of CT. Dr John C. Dickson, Principal Physicist UCLH

Algebraic Iterative Methods for Computed Tomography

Investigation on reconstruction methods applied to 3D terahertz computed Tomography

Learning Splines for Sparse Tomographic Reconstruction. Elham Sakhaee and Alireza Entezari University of Florida

CIVA CT, an advanced simulation platform for NDT

Joint ICTP-TWAS Workshop on Portable X-ray Analytical Instruments for Cultural Heritage. 29 April - 3 May, 2013

Tomography at all Scales. Uccle, 7 April 2014

NIH Public Access Author Manuscript Med Phys. Author manuscript; available in PMC 2009 March 13.

Medical Image Reconstruction Term II 2012 Topic 6: Tomography

Compressed Sensing for Electron Tomography

Tomographic Reconstruction

Quality control phantoms and protocol for a tomography system

Computed Tomography for Industry Needs and Status Umesh Kumar

Generalized Filtered Backprojection for Digital Breast Tomosynthesis Reconstruction

Introduction to Positron Emission Tomography

BME I5000: Biomedical Imaging

Introduction to Biomedical Imaging

Background. Outline. Radiographic Tomosynthesis: Image Quality and Artifacts Reduction 1 / GE /

Spiral ASSR Std p = 1.0. Spiral EPBP Std. 256 slices (0/300) Kachelrieß et al., Med. Phys. 31(6): , 2004

Ch. 4 Physical Principles of CT

Algebraic Iterative Methods for Computed Tomography

Multilevel Optimization for Multi-Modal X-ray Data Analysis

Workshop on Quantitative SPECT and PET Brain Studies January, 2013 PUCRS, Porto Alegre, Brasil Corrections in SPECT and PET

GPU-Based Acceleration for CT Image Reconstruction

Temperature Distribution Measurement Based on ML-EM Method Using Enclosed Acoustic CT System

Interior Reconstruction Using the Truncated Hilbert Transform via Singular Value Decomposition

X-ray Tomography. A superficial introduction, but sufficient enough to get us started in surgical navigation.

DEVELOPMENT OF CONE BEAM TOMOGRAPHIC RECONSTRUCTION SOFTWARE MODULE

REDUCED ORDER MODELING IN MULTISPECTRAL PHOTOACOUSTIC TOMOGRAPHY

Tomographic reconstruction: the challenge of dark information. S. Roux

Total Variation and Tomographic Imaging from Projections

Computer-Tomography II: Image reconstruction and applications

Multi-slice CT Image Reconstruction Jiang Hsieh, Ph.D.

Reconstruction of CT Images from Sparse-View Polyenergetic Data Using Total Variation Minimization

Enhanced material contrast by dual-energy microct imaging

Index. aliasing artifacts and noise in CT images, 200 measurement of projection data, nondiffracting

Recognition and Measurement of Small Defects in ICT Testing

CoE4TN4 Image Processing. Chapter 5 Image Restoration and Reconstruction

Radiology. Marta Anguiano Millán. Departamento de Física Atómica, Molecular y Nuclear Facultad de Ciencias. Universidad de Granada

arxiv: v2 [cond-mat.mtrl-sci] 5 Jan 2010

HIGH-SPEED THEE-DIMENSIONAL TOMOGRAPHIC IMAGING OF FRAGMENTS AND PRECISE STATISTICS FROM AN AUTOMATED ANALYSIS

Deformation of granular texture media studied by X-ray CT & 3D DIC at the continuous and microstructure scales

Phase-Contrast Imaging and Tomography at 60 kev using a Conventional X-ray Tube

Development of a multi-axis X-ray CT for highly accurate inspection of electronic devices

A Projection Access Scheme for Iterative Reconstruction Based on the Golden Section

Review of PET Physics. Timothy Turkington, Ph.D. Radiology and Medical Physics Duke University Durham, North Carolina, USA

Limited View Angle Iterative CT Reconstruction

Iterative CT Reconstruction Using Curvelet-Based Regularization

Full-Colour, Computational Ghost Video. Miles Padgett Kelvin Chair of Natural Philosophy

NON-COLLIMATED SCATTERED RADIATION TOMOGRAPHY

GE s Revolution CT MATLAB III: CT. Kathleen Chen March 20, 2018

Dr. Javier Santillan, San Carlos, CA

Axial block coordinate descent (ABCD) algorithm for X-ray CT image reconstruction

Biomedical Imaging. Computed Tomography. Patrícia Figueiredo IST

Moscow-Bavarian Joint Advanced Student School 2006 / Medical Imaging Principles of Computerized Tomographic Imaging and Cone-Beam Reconstruction

DUE to beam polychromacity in CT and the energy dependence

3D TeraHertz Tomography

A new calibration-free beam hardening reduction method for industrial CT

Financial disclosure. Onboard imaging modality for IGRT

Biophysical Techniques (BPHS 4090/PHYS 5800)

SYSTEM LINEARITY LAB MANUAL: 2 Modifications for P551 Fall 2013 Medical Physics Laboratory

Digital Image Processing

The Electrochemical Innovation Lab X-ray Suite: from macro- to nano-ct

Image Reconstruction from Projection

Reconstruction from Projections

Digital Volume Correlation for Materials Characterization

Total Variation Regularization Method for 3D Rotational Coronary Angiography

Fast Reconstruction of CFRP X-ray Images based on a Neural Network Filtered Backprojection Approach

Image Acquisition Systems

Reconstruction Methods for Coplanar Translational Laminography Applications

Introduction to Emission Tomography

Estimating 3D Respiratory Motion from Orbiting Views

UNIVERSITY OF SOUTHAMPTON

Cover Page. The handle holds various files of this Leiden University dissertation

Introduc)on to PET Image Reconstruc)on. Tomographic Imaging. Projec)on Imaging. Types of imaging systems

MULTI-PURPOSE 3D COMPUTED TOMOGRAPHY SYSTEM

SPECT reconstruction

SUPPLEMENTARY INFORMATION

LABORATORY SYSTEM FOR X-RAY NANOTOMOGRAPHY

Empirical cupping correction: A first-order raw data precorrection for cone-beam computed tomography

MEDICAL IMAGING 2nd Part Computed Tomography

A Curvelet based Sinogram Correction Method for Metal Artifact Reduction

Sparse Reconstruction / Compressive Sensing

Expectation Maximization and Total Variation Based Model for Computed Tomography Reconstruction from Undersampled Data

CT Systems and their standards

8/7/2017. Disclosures. MECT Systems Overview and Quantitative Opportunities. Overview. Computed Tomography (CT) CT Numbers. Polyenergetic Acquisition

Research Article SART-Type Image Reconstruction from Overlapped Projections

Central Slice Theorem

Image Reconstruction in Optical Tomography : Utilizing Large Data Sets

Introduction to Topics in Machine Learning

Transcription:

Méthodes d imagerie pour les écoulements et le CND Journée scientifique FED3G CEA LIST/Lab Imagerie Tomographie et Traitement Samuel Legoupil 15 juin 2012

2D/3D imaging tomography Example Petrochemical reactor scanner : Liquid distribution in a fixed bed reactor (3 hydrodynamics conditions) µtomography imaging : 3D-Fuel injector Ni foam for gas distributor and 3D image analysis results Carbon GDL For fuel cell (fiber=5µm)l Menisci shape in A 100x20 µm canal For microfluidics XPIV for flow characterization 2

Outline Context Basics of CT and industrial constraints Projection matrix and modeling needs CS reconstruction methods Implémentation and adaptative data representation Future developments 3

Introduction Tomographic Imaging Reconstruction with Non-diffracting Sources Tomographic Imaging with Diffracting Sources Reflection Tomography 4

Tomography under progress Number of papers with x-ray and CT, From Ge Wang - Med. Phys. 35, March 2008 5

Differences between medical and industrial imaging Medical Industrial Sample dimensions =0.75 x L=2 m² 10-3 3m Sample dynamics 0 50 Hz 0 10 khz Dose As low as possible No constraint Processing time Short No constraint Contrast 1% 1% Sampling conditions Good Weak most oftenly partial view of object 6

Context Growing-up of X-ray CT controls as a standard tool: Material science, biology, Earth science, archeology CT integration for on-line control of manufactured objects turbine blades, medicine industry for spray dispenser High complexity of the method Tedious experimental optimization NIKON 7

Outline Context Basics and industrial constraints Projection matrix and modeling needs CS reconstruction methods Implementation and adaptative data representation Future developments 8

CT General principles Physics Absorption of media, depends on : Material characteristics (Z density, density, thickness) Photon energy E I I 0 e ( E) d d The optimal energy for the measurement must satisfy : ( E) 2 d 9

Absorption coefficient (cm²/g) CT General principles Optimal energy 100 µ - Xcom (cm²/g) ( E) c E d 10 µ - Fit (Power) 5 E 40keV 1 0.1 0 10 20 30 40 50 Energy (kev) 10

Sample thickness d (cm) Optimal energy (kev) Relative error on measurement CT General principles Optimal energy E d ( E) 2 d 1.2 1 0.8 R=a/b=10 18 16 14 12 100 10 E=6keV E=16keV E=11keV 10 0.6 8 1 0.4 6 4 0.1 0.2 2 0 0 0 15 30 45 60 75 90 105 120 135 150 165 180 Projection angle 0.01 0 15 30 45 60 75 90 105 120 135 150 165 180 Projection angle 11

Constraints on acquisition Projection configurations 12

Constraints on object Object configuration Partial view of the object Complex support of reconstruction domain 13

Outline Context Basics and industrial constraints Projection matrix and modeling needs CS reconstruction methods Implementation and adaptative data representation Future developments 14

Iterative Algorithms For example, ML-EM, OS-EM, ART, GC, Discretize the image into pixels Solve imaging equations AX=P P X = unknowns (pixel values) P = projection data A = imaging system matrix a ij X 15

Iterative Algorithms Regularisation - example Noise model Data matching Prior encouragement V F i 1 2 ( A ) ( 2 i X pi V x i i x j ) If V(x) = x 2, it enforces smoothness. 16

Inversion methods Statistical approach Problem description in Emission Tomography Problem description in Transmission Tomography (CT) Estimation of H Law of noise Choice of approach and associated algorithm(s) Iterative method : Statistical inversion : fˆ fˆ arg max x arg max P( g x f ) P( g f ) P( f ) 17

Projection matrix in algebraic approaches Reconstruction model I I( E)exp( µ ( x, E) dx) de I I 0 exp( µdx) i Voxel y ln I I 0 µ j A i j Aµ z y x 18

Projection matrix in algebraic approaches y j J y Aµ 2 a ij Projection (A): y j a ij i Backprojection (A T ): i j a ij y i j Source 19

Photons scattering in cone beam CT Evolution of build-up (1+N scattered photons /N direct photons ) D source-object =50 mm D source-detector =200 mm Det. size=100 mm U=160 kv 20

Projection matrix in algebraic approaches y j J y Aµ 2 a ij Projection (A): y j a ij i i a i j Distance to beam axis Backprojection (A T ): i j a ij y j Source 21

Projection matrix in algebraic approaches Example of weighted function calculated with CIVA 22

Outline Context Basics and industrial constraints Projection matrix and modeling needs CS reconstruction methods Implementation and adaptative data representation Future developments 23

Reconstruction from few projections The objective is the reconstruction from few X-ray projections: Speed-up of acquisition process Low-dose for medical imaging The inverse problem is highly ill-posed need for specific approach Original object Frequential domain projections (11 proj.) Reconstruction by analytic inversion 24

Reconstruction from few projections Recent developed Compressed sensing theory confirms that with a sampling rate lower than the Nyquist rate, we can also perfectly reconstruct a signal. - Sparse representation of signal under appropriate basis (sinusoid, wavelet ) - Reconstruction solving a convex optimization problem. min x x 1 s. t. Ax b 2 25

Reconstruction from few projections As images are rarely sparse, the idea is to find a transform α=(µ) such that the transformed coefficients are sparse, eg Total Variation of µ: TV( ) N 1 N 1 i0 j0 x 2 i, j y 2 i, j min s. t. 1 * A b 2 * Original object Reconstruction by analytical inversion Compressed sensing 26

Methods comparison Reconstruction from 32 projections FBP OS-EM Comp. Sensing 27

Outline Context Basics and industrial constraints Projection matrix and modeling needs CS reconstruction methods Implementation and adaptative data representation Future developments 28

Computer implementation Computer implementation has to estimate g Hf and f H T g Reconstruction Volume = N 3 voxels 2D detector=n² pixels N p projections Matrix H sizes N p xn 2 xn (non-null elements) N N p GBytes 256 64 2 512 256 64 1024 512 1024 2048 512 8192 GPU hardware 29

Example: Ni foam drying ART Bayesian approach Speed-up factors: x 240 on projection x 80 on back projection Reconstruction volumes : 1024 3 2048 3 31

Adaptative information representation 32

Adaptative information representation 33

CIVA NDT Platform CT module 34

2D knee reconstruction 35

Outline Context Basics and industrial constraints Projection matrix and modeling needs CS reconstruction methods Implementation and adaptative data representation Future developments 36

Instantenous imaging Synchronous phenomenon Asynchronous phenomenon max =25% Pompe axis max =39% Relevant information max =44% Multi sources Multi sensors 2 juillet 2008 37

Nouvelles technologies de sources X Avantages des sources à base de CNT : Fort courant (100 µa/tube) Source nano-foyer Très haute résolution spatiale Imagerie par contraste de phase Source étendue Haute intensité Codage de source Sources multiples Tomographie dynamique D. Pribat Voir http://www.nanosprint.com/index.php?id= 2 juillet 2008 38

Simulation of robot CT Local imaging on a «C3» : conformity assessment Defect detection Competition analysis Reconstructed «middle foot 40

CIVA-RT / CIVA-CT Platform for NDT evaluation (US, CF and X-ray) GUI Tomo Analytical algo CIVA RX Civa visualisation Plug-in Tomo Algebraic algo Statistical algo High performance calculation 41

Future developments Industrial situation Modeling Data Processing Information Sensors Data acquisition Hardware developments X-ray detector: fine resolution, dynamic, multi energy X-ray generator: microfocus source (@ 200 nm), adaptative, c Modeling capacities (from physics to signal theory) and reconstruction Database reconstruction Computation capacities (GPU hardware, Larabi ) Information processing 42

43