(Tuesday, 7/31/2018) 10:00 AM - 10:30 AM
Room: Exhibit Hall | Forum 1
Purpose: Advanced CBCT Iterative Reconstruction (IR) algorithms for scatter correction, motion compensation, noise suppression, etc. have been developed for image guided radiotherapy (IGRT). But the clinical implementation is hindered by their high computational cost. We developed IR algorithms in polar pixelation grid that can dramatically improve computational efficiency by taking advantage of the rotational symmetry of the CT geometry. In this work, we characterized the impacts of heterogeneous pixelation on image quality.
Methods: The images are discretized in a polar grid with azimuthal divisions matching the projection angles. The system matrices for both traditional Cartesian and polar IR were calculated using a numerical method and stored. The images were reconstructed by Simultaneous Algebraic Reconstruction Technique (SART) with pre-computed system matrices as well as filtered back projection (FBP) algorithms. XCAT2 numerical phantom was used to generate projection data with added noises. Three figures of merit (FOM) referring to the original image were employed to quantify the noise levels and reconstruction accuracy. The modulation transfer functions (MTF) were calculated to evaluate image resolution.
Results: Image resolution was compromised if projection number is insufficient, but can be recovered by further dividing the voxels of outer rings. The image noise distribution was affected by the polar to Cartesian image conversion method. Linear interpolation method generates more noises in the image center while the area mapping method can produce a uniform noises similar to Cartesian IR.
Conclusion: Despite of its advantage in computation efficiency, the polar IR algorithm needs to be carefully designed to achieve optimal performance. Optimized polar IR algorithms may achieve similar image quality as traditional Cartesian IR.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
