MENTORS

Prof. Gabor Fichtinger, Csaba Pinter, Andras Lasso

info: www.cs.queensu.ca/~gabor, email: gabor@cs.queensu.ca

BACKGROUND AND SIGNIFICANCE

Segmentation results (contours, structures) can be represented in various ways. In the medical image standard DICOM, they are stored as a series of 2D contours (ribbons), but most analysis and processing algorithms require binary volumes (labelmaps) as input. Thus, it is crucial to have a robust, stable, and accurate way to do the conversion, which is called rasterization.

The radiation therapy toolkits need to deal with structure sets as part of their workflows, so they typically contain a conversion algorithm. So do the SlicerRT (developed in the PerkLab) and the CERR research toolkits, and most commercial treatment planning systems. However, the assumptions and the algorithms differ, thus the processing results differ as well. It is needed to evaluate and compare the available rasterization methods, and find out the effect of certain options that can be changed in such an algorithm.

PROJECT OBJECTIVES

Familiarize with the rasterization algorithm of SlicerRT and CERR, create a report on the similarities and differences. Tweak the SlicerRT rasterization method and see what effect the certain decisions have on the result labelmap. Compare the result dose volume histogram (DVH) curves created using CERR and the different SlicerRT variations, as well as the available commercial software (such as Eclipse used in KGH). Find the optimal rasterization setting, expose the necessary options and evaluate the optimal algorithm.

SKILLS DEVELOPED

Context: Image processing algorithm evaluation.

Analytical: Image processing algorithm.

Experimental: Algorithm evaluation and development, using the 3D Slicer platform.

BACKGROUND REQUIRED

Practical aptitude and mindset, C++ programming, good Matlab skills.

Preferred prerequisites: Any combination of Computer Integrated Surgery, Computer Graphics, Medical Informatics or Medical Imaging.