Segmentation of contrast enhanced CT images for attenuation correction of PET/CT data

Abstract

The use of contrast media in PET/CT imaging has been suggested to cause PET artifacts during the CT-based attenuation correction process. In this paper, we evaluate three algorithms that segment intravenous (IV) contrast-enhanced tissue from chest CT images to minimize possible artifacts. The algorithms that were evaluated are template matching, 3D region growing, and snake-based technique. These methods were tested using 5 patient studies. The segmentation result of each method was compared to its corresponding manually segmented images on a voxel-wise basis, and a squared difference between the two segmentation results was calculated. The averaged squared differences of all 5 patients for the template matching, region growing, and snake-based method were 19.0%plusmn7.1%, 65.2%plusmn51.5%, and 13.5%plusmn6.5% respectively. We concluded that the snake model is most suitable for efficiently segmenting the contrast-enhanced CT images among the three methods