PVEOut: Implementation of pvelab a framework for correcting PET and SPECT data for Partial Volume Effects PVEOut Satellite meeting Budapest, June 11 th 2004 Tim Dyrby, Thomas Rask & Claus Svarer, NRU Marco Comerci & Mario Quarantelli, IBB

NRU, 2004 Background In the PVEOut research program one of the task was that a commercial and suitable software interface for the PVE correction process should be developed and implemented so it was possible to use the PVE correction routines by research groups and for clinical purposes Original it was agreed that: –The research groups involved in the PVEOut project should implement their routines using the methods and programming language that they preferred –The company should then transfer the methods to a common framework, including a common graphical interface so it was easy to use and very stable

NRU, 2004 Background When the company withdrew from the project it was decided that the groups would try to implement this common user interface for the PVE correction routines The research groups in the PVEOut project consist of: –Biostructure and Bioimaging Institute (IBB), Naples (coordinator) –Neurobiology Research Unit, Rigshospitalet, Copenhagen –Dept. of clinical Neuroscience, Karolinska Institute, Stockholm –University of Kent, Canterbury –INSERM, U320, Caen –PET centre, University of Debrechen, Debrechen these groups have all contributed to the development of parts of the pvelab software

NRU, 2004 Background Opposite the original idea where all software should be re- implemented in C++ using X-Windows for visualization, it was chosen to use a more open implementation so no re- implementation of the routines for performing each step in PVE correction process is needed. The following decision were made about the implementation: –The graphical user interface to the software will be implemented in Matlab (easier as the groups had expertise in this) –It should be possible to incorporates routines implemented the way they already are implemented (in C, C++, Fortran, and Matlab) –The framework should be able to run at a Linux workstation with Matlab 6.5 (preferably it should also be able to run on Windows systems and other UNIX variants) Thereafter we were starting designing PVElab

NRU, 2004 Design issues for pvelab The software should be able to: –automate the PVE correction process –control the performance of each step in the image processing pipeline –give feed-back about performance –visualize results during processing –log results of each step in the process for later inspection and very essential: it should possible (and preferably very easy) to include own methods for each step in the PVE correction process

NRU, 2004 Timeline for design and implementation of pvelab October 2002: The company withdrew from the project January 2003: First version of analysis and design was presented for the PVEOut project group by Tim Dyrby, NRU July 2003: Design finished and an almost finished PVELab interface implementation was presented October 2003: Tested implementation of the task interface presented January 2004: Task wrappers finished and the total PVE correction process up and running, Thomas Rask, NRU and Marco Comerci, IBB April 2004: Minor modifications to task interfaces implemented and Beta release note send out

NRU, 2004 pvelab: Main interface Implemented in Matlab: Tab for each task in the process

NRU, 2004 pvelab: Implemented tasks pvelab: software for control and logging of the image processing pipeline (NRU) Fileload: Software for loading and converting analyze and dicom image files Registration: Interactive Image Overlay (IIO, NRU) Interactive Point Selection (IPS, NRU) Interface to SPM co-registration (Statistical Parametric Mapping) Interface to AIR registration (Automatic Image Registration) Load AIR file (registration done otherwise) Segmentation: QMCI segmentation (Naples, uses 3 T1, T2, and PD) Interface to SPM segmentation (Statistical Parametric Mapping) Interface to BrainSeg segmentation (Canterbury) Load segmentation volume (segmentation done otherwise)

NRU, 2004 pvelab: Implemented methods Reslice: Interface to ResliceWarp (using Brain Warp, Kjems, DTU) Interface to AIR reslicing (Automatic Image Registration) Reslice (pure matlab, really time and memory consuming) Atlas (defines VOI’s (volumes of interest): Talairach based (Naples, 16 VOI’s) MNI based (Naples, requires T1, T2 and PD images) applyrois (NRU, 37 VOI’s) PVE correction: PVE correction (Naples) Further: Programs for visualization of images (NRU) Programs for correction of analyze files (headers) Other tools for correcting images

NRU, 2004 pvelab: More possibilities For each step it is really easy to incorporate an interface for your own favorite method and the framework can be used for setting up a pipeline for automating other image processing tasks

NRU, 2004 pvelab: Menus & tools

NRU, 2004 pvelab: Options for tasks All possibilities for each of the implemented methods is available by selecting options.

NRU, 2004 pvelab: Registration & Show Possible methods: Interactive Image Overlay (IIO) Interactive Point Selection (IPS) Interface to SPM (co-registration) Interface to AIR (alignlinear) Possibility for inspecting results within pvelab:

NRU, 2004 pvelab: View -> Inspecting results of registration Controlling registration more carefully using: pvelab -> View -> Inspect

NRU, 2004 pvelab: Segmentation

NRU, 2004 pvelab: PVE correction For PVE correction the most common methods are available and reported in the output

NRU, 2004 pvelab: How do you actually perform a PVE correction

NRU, 2004 Pvelab: How do you get it A possibility for setting up a PVE correction pipeline without having to program yourself Have already been downloaded by several different centers (Beta release status) Can be download’ed from: – –remember also to download: mriwarp-1.03.tar.gz and compile it for your operating system –and Matlab 6.5, Automatic Image Registration (AIR 5), and Statistical Parametric Mapping (spm2) are required. A new version will be out beginning of July. Register sending an to and we will contact you.