Validation of An Iterative Deconvolution Algorithm for the Post Treatment Y90 PET Image of Liver Cancer
H Qu*, N Yu, P Xia, The Cleveland Clinic Foundation, Cleveland, OHSU-E-J-189 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
To validate the interactive deconvolution method for dose calculation from breathing effect removed Y90 PET image and to quantify impact of the breathing pattern variation on the dose distribution
A computer phantom simulating the PET image of a liver tumor was constructed to investigate an internally developed program using the Van Cittert of deconvolution to remove the breathing motion effect from PET images. The tumor radius was 4 cm. Typical pixel values and noise inside and outside the tumor were taken from real patient Y90 PET images (inside tumor 415±164, outside tumor 15 ±37, negative pixels were clipped to 0). The phantom PET image was convolved with a breathing kernel (breathing amplitude 2 cm). These images were then deconvolved with the Van-Cittert deconvolution. The pixel values of the convolved and the deconvolved image are compared with the original image.
Results:The root mean square (rms) of the pixel difference between the original and convoluted image was 3.33, and between the original and the deconvolved images was 0.54.
Conclusion:The deconvolution removed most of the changes in pixel value sue to breathing.