Improving PET/CT attenuation correction with iterative CT beam hardening correction

Due to x-ray polychromaticity CT attenuation data cannot be directly used for PET attenuation correction (AC). Usually, the polychromatic CT projection data are converted to the desired 511 keV by image-based image domain scaling methods. However, due to the nonlinearity of attenuation this is a zero order correction only. Ideally, corrective measures must be applied in the rawdata domain. We therefore implemented a beam hardening correction (BHC) algorithm dedicated for PET/CT. Our method is image-based since it starts from CT images and provides corrected CT images. Internally it operates in both the image domain and the projection data domain. The algorithm resamples the CT volume to PET resolution and performs a weighted thresholding to obtain a material distribution. Then, the set of material images undergoes an iterative correction process that consists of several forward projections and backprojections along the original ray direction. Although this can be done until convergence we found that the material images do not suffer from beam hardening artifacts anymore after just one iteration. From these corrected material distributions the desired monochromatic CT image at 511 keV is easily obtained by linear combination. A simulation study, phantom and patient measurements with a Sensation 4 and a Sensation 64 spiral multi-slice CT scanner (Siemens Medical Solutions, Forchheim, Germany) were carried out to compare the 511 keV volume obtained with iBHC to the standard image-based method. In the presence of dark streaks due to beam hardening, for example in the head or pelvis region, the iterative approach achieves to remove these streaks whereas the image-based scaling cannot do so. The scaled images show differences of up to 50 HU in the pelvis region. The additional computation time required lies in the order of 0.2 s per slice and material. Our results indicate significant improvements over the standard approach. However, a clinical evaluation with PET/CT data must follow to finally validate our results.