Statistical Image Reconstruction Methods

 

Short Course Organizer and Instructors

 

·            Dr. Freek Beekman

·            Associate professor responsible for R&D on Tomography

·            Image Sciences Institute, University Medical Centre Utrecht

·            +31 30 2538843

·            +31 30 2539032

·            Freek Beekman (physicist, Ph.D.’95, associate professor at the Image Science Institute, Utrecht University, authored more than 40 journal peer reviewed journal papers, several book chapters and patent applications. His research interests include image reconstruction (in particular emission CT and X-ray CT), Monte Carlo and analytic modelling, and tomographic instrumentation.

 

 

·            Johan Nuyts

·            Associate professor, responsible for R&D in nuclear medicine imaging.

·            Affiliation: Dept. Nuclear Medicine, K.U.Leuven, Belgium.

Address: Nuclear Medicine, UZ Gasthuisberg, Herestraat 49, B3000 Leuven, Belgium.

·            Tel: +32 16 343715

·            Fax: +32 16 343759

·            E-mail: Johan.Nuyts@uz.kuleuven.ac.be

·            Degree of Electronical Engineering in ’82, of Medical Physics in ’83 and PhD in Applied Sciences in ’91. R&D in image processing hardware until ’87 at ESAT, MI2, K.U.Leuven, and since ’87 R&D in nuclear medicine imaging, UZ Gasthuisberg and K.U.Leuven. His research interests include iterative reconstruction and image processing for nuclear medicine applications.

 

 

 

Course Abstract

·       Statistical Image Reconstruction Methods

·       Description of Course

The introduction of novel PET, SPECT and CT imaging devices, availability of fast computers and algorithms, as well as the increasing demand for improved image reconstruction has brought new relevance to the topic of discrete reconstruction methods. These include methods that are suitable for modelling noise in the projection data, for incorporating prior knowledge about the object to be reconstructed, and for model-based correction of image degrading effects (i.e. detector blurring, photon attenuation and scatter).

 

The objective of this four hours course is to provide up-to-date practical knowledge on the emerging area of discrete image reconstruction, applied to SPECT, PET, and transmission CT. The course will cover topics like scatter correction for emission tomography and CT, beam-hardening correction for X-ray CT, resolution recovery through the modelling of blurring, noise suppression through Bayesian methods and post-filtering, and characteristics of different types of algorithms applied to different modalities. In all cases, numerous examples will be presented of the present state-of-the-art.

 

Prequisite knowledge should include basics of the physics of imaging systems, statistics, and elementary linear algebra.

 

COURSE OUTLINE

 

I. Introduction and primer (Freek Beekman)

A. Imaging modalities and their discrete models

B. Iterative methods

C. Primer on block-iterative and dual matrix methods

D. Primer on Regularization

 

II. Theory (Johan Nuyts)

A. Maximum likelihood reconstructions

B. Bayesian reconstruction

 

III. Modelling of photon transport (Freek Beekman)  

A. SPECT: attenuation, scatter and blurring correction.

B. PET:   attenuation, scatter and blurring correction

C. Beam hardening and scatter modelling for X-ray CT

D. Resolution recovery in transmission CT

 

IV. Capita Selecta

A. Using the Fisher information matrix in iterative reconstruction (Johan Nuyts)

B. Monte Carlo modelling for quantitative SPECT and CT (Freek Beekman)

 

 

V  Various practical implementation tricks

     (Spike suppression through pre-processing of images, blurring strategies for resolution recovery, filter versus Bayesian, etc, etc)