GPU-based Image Processing Methods in Higher Dimensions and their Application to Tomography Reconstruction Szirmay-Kalos, László Budapest Uni of Tech Sapporo,

Slides:



Advertisements
Similar presentations
SHREYAS PARNERKAR. Motivation Texture analysis is important in many applications of computer image analysis for classification or segmentation of images.
Advertisements

Fast Approximation of Multiple Scattering in Inhomogeneous Participating Media Szirmay-Kalos László Liktor Gábor Tamás Umenhoffer Tóth Balázs EG 2009.
Prepared 7/28/2011 by T. O’Neil for 3460:677, Fall 2011, The University of Akron.
Chapter 8 Planar Scintigaraphy
Szirmay-Kalos László Magdics Milán Tóth Balázs
Approaches for Retinex and Their Relations Yu Du March 14, 2002.
Implementing the Probability Matrix Technique for Positron Emission Tomography By: Chris Markson Student Adviser: Dr. Kaufman.
Lecture 19 Continuous Problems: Backus-Gilbert Theory and Radon’s Problem.
Part 4 b Forward-Backward Algorithm & Viterbi Algorithm CSE717, SPRING 2008 CUBS, Univ at Buffalo.
Back-Projection on GPU: Improving the Performance Wenlay “Esther” Wei Advisor: Jeff Fessler Mentor: Yong Long April 29, 2010.
Apparent Emissivity in the Base of a Cone Cosmin DAN, Gilbert DE MEY University of Ghent Belgium.
Rician Noise Removal in Diffusion Tensor MRI
Maurizio Conti, Siemens Molecular Imaging, Knoxville, Tennessee, USA
Basic Principles of Computed Tomography Dr. Kazuhiko HAMAMOTO Dept. of Infor. Media Tech. Tokai University.
Presentation by Dr. David Cline Oklahoma State University
PET data preprocessing and alternative image reconstruction strategies.
Introduction to Adaptive Digital Filters Algorithms
TeraTomo project: a fully 3D GPU based reconstruction code for exploiting the imaging capability of the NanoPET™/CT system M. Magdics 1 ), L. Szirmay-Kalos.
Design and simulation of micro-SPECT: A small animal imaging system Freek Beekman and Brendan Vastenhouw Section tomographic reconstruction and instrumentation.
A flexible FGPA based Data Acquisition Module for a High Resolution PET Camera Abdelkader Bousselham, Attila Hidvégi, Clyde Robson, Peter Ojala and Christian.
Cg Programming Mapping Computational Concepts to GPUs.
Optimizing Katsevich Image Reconstruction Algorithm on Multicore Processors Eric FontaineGeorgiaTech Hsien-Hsin LeeGeorgiaTech.
EE369C Final Project: Accelerated Flip Angle Sequences Jan 9, 2012 Jason Su.
Particle Filters for Shape Correspondence Presenter: Jingting Zeng.
Medical Image Analysis Image Reconstruction Figures come from the textbook: Medical Image Analysis, by Atam P. Dhawan, IEEE Press, 2003.
02/10/03© 2003 University of Wisconsin Last Time Participating Media Assignment 2 –A solution program now exists, so you can preview what your solution.
Volume radiosity Michal Roušal University of West Bohemia, Plzeň Czech republic.
Image Reconstruction from Projections Antti Tuomas Jalava Jaime Garrido Ceca.
Development of a Baseline Tropical Cyclone Model Using the Alopex Algorithm Robert DeMaria.
Path Integral Formulation of Light Transport
Today’s lecture 2-Dimensional indexing Color Format Thread Synchronization within for- loops Shared Memory Tiling Review example programs Using Printf.
Dense Image Over-segmentation on a GPU Alex Rodionov 4/24/2009.
Parallel Algorithms Patrick Cozzi University of Pennsylvania CIS Spring 2012.
Professor Brian F Hutton Institute of Nuclear Medicine University College London Emission Tomography Principles and Reconstruction.
75 th Annual Meeting March 2011 Imaging with, spatial resolution of, and plans for upgrading a minimal prototype muon tomography station J. LOCKE, W. BITTNER,
O AK R IDGE N ATIONAL LABORATORY U.S. DEPARTMENT OF ENERGY Image Reconstruction of Restraint-Free Small Animals with Parallel and Multipinhole Collimation:
1 Markov random field: A brief introduction (2) Tzu-Cheng Jen Institute of Electronics, NCTU
Nuclear Medicine: Tomographic Imaging – SPECT, SPECT-CT and PET-CT Katrina Cockburn Nuclear Medicine Physicist.
Iteration Solution of the Global Illumination Problem László Szirmay-Kalos.
Arithmetic Test Pattern Generation: A Bit Level Formulation of the Optimization Problem S. Manich, L. García and J. Figueras.
Expectation-Maximization (EM) Algorithm & Monte Carlo Sampling for Inference and Approximation.
Introduction In positron emission tomography (PET), each line of response (LOR) has a different sensitivity due to the scanner's geometry and detector.
Motion Estimation using Markov Random Fields Hrvoje Bogunović Image Processing Group Faculty of Electrical Engineering and Computing University of Zagreb.
CS/EE 217 GPU Architecture and Parallel Programming Midterm Review
Introduction In Positron Emission Tomography (PET), each line of response (LOR) has a different sensitivity due to the scanner's geometry and the detector's.
Gamma Photon Transport on the GPU for PET László Szirmay-Kalos, Balázs Tóth, Milán Magdics, Dávid Légrády, Anton Penzov Budapest University of Technology.
Statistical Methods for Image Reconstruction
Error Diffusion (ED) Li Yang Campus Norrköping (ITN), University of Linköping.
Air Systems Division Definition of anisotropic denoising operators via sectional curvature Stanley Durrleman September 19, 2006.
PATH INTEGRAL FORMULATION OF LIGHT TRANSPORT Jaroslav Křivánek Charles University in Prague
Strong Supervision From Weak Annotation Interactive Training of Deformable Part Models ICCV /05/23.
Computer graphics III – Low-discrepancy sequences and quasi-Monte Carlo methods Jaroslav Křivánek, MFF UK
Density Estimation in R Ha Le and Nikolaos Sarafianos COSC 7362 – Advanced Machine Learning Professor: Dr. Christoph F. Eick 1.
RECONSTRUCTION OF MULTI- SPECTRAL IMAGES USING MAP Gaurav.
Generalization Performance of Exchange Monte Carlo Method for Normal Mixture Models Kenji Nagata, Sumio Watanabe Tokyo Institute of Technology.
mps-tk : A C++ toolkit for multiple-point simulation
Leiming Yu, Fanny Nina-Paravecino, David Kaeli, Qianqian Fang
Probabilistic Models for Linear Regression
Single Photon Emission Tomography
Tianfang Li Quantitative Reconstruction for Brain SPECT with Fan-Beam Collimators Nov. 24th, 2003 SPECT system: * Non-uniform attenuation Detector.
Lecture 5 – Improved Monte Carlo methods in finance: lab
Basic Principles of Computed Tomography
CS/EE 217 – GPU Architecture and Parallel Programming
Parallelizing the Condensation Algorithm for Visual Tracking
Mont-Carlo simulation of OCT structural images of subcutaneous
Iterative Algorithm g = Af g : Vector of projection data
Non-local Means Filtering
Monte Carlo Rendering Central theme is sampling:
Ray Tracing - Analysis of Super-Sampling Methods
Presentation transcript:

GPU-based Image Processing Methods in Higher Dimensions and their Application to Tomography Reconstruction Szirmay-Kalos, László Budapest Uni of Tech Sapporo, 2010

Positron Emission Tomography e-e- e+e+ Line Of Response: y Intensity: x

Iterative Maximum Likelihood Reconstruction Measured detector response Source intensity as a 3D voxel array Source estimation Source correction Compute expected detector response Expected detector response

Ill-posed reconstruction error Iteration number Maximum likelihood estimate

Regularization Additional information –Penalty term added to the likelihood Prevents overfitting TV norm (L1 optimization) –No smoothness condition –Preserves edges

TV minimalization In steepest descent search the derivative of the TV term is needed: –Function |x| cannot be differentiated: Add a small term (blurring) Primal-dual methods –Only local values are needed: parallelization xVxV

Detector scattering compensation Path probability inside the detector can be pre-computed or measured photon crystals intercrystal scattering absorption Electronics number of hits

Pre-computation 

L w L = Quasi-Monte Carlo filtering

Random sampling undersampling oversampling

Delta-Sigma modulator Filter kernel pixels

Filter kernel Delta-Sigma modulator

Filter kernel Floyd-Steinberg halftoning!

Sampling with Sigma-Delta modulation

GPU Implementation Simulation step: GPU: Quasi-SIMD massively parallel machine –Gathering = threads to equations (outputs) –“No” conditional statements or variable length loops Reconstruction algorithm –Geometric LOR marching: threads to LORs (adjoint problem) –LOR filtering: threads to output LORs –TV regularization: threads to voxels high dim. integrals 10 8 voxels 10 8 LORs

TV regularization results =0.005 =0.05 =0.008 No TV

TV results =0.001 = = =0.005

Scattering in the detector 3D reconstruction, no detector scattering compensation Detector scattering compensation 2D reconstruction: SSRB + OSEM

F18 mouse

Conclusions Image processing algorithms can be and are worth being generalized to higher dimensions, but beware the curse of dimensions and use Monte Carlo methods. GPUs are good platforms for image processing, but adopt the gathering view and refrain from conditionals.