Reduction of Intravenous Contrast Related Artifacts in CT-Based Attenuation Corrected PET Images

A well established advantage of using contrast-enhanced CT in dual-modality PET/CT imaging is the possibility to obtain diagnostic quality CT images, thus allowing to increase patient throughput and reduce patient absorbed dose. However, the use of CT data following intravenous (IV) injection of contrast media for attenuation correction might bias the attenuation-map (mu map) and generate visible artifacts on reconstructed PET images owing to the misclassification of contrast-medium with high density bone when using standard procedures for conversion of CT numbers to linear attenuation coefficients at 511 keV. In this study, an algorithm for segmentation and classification of irregular shapes of regions containing IV contrast-medium usually found in clinical CT images is proposed and assessed using clinical data. The proposed automated three-dimensional segmentation method consists of two steps: firstly all objects with high CT number are segmented based on a region-based segmentation method in conjunction with boundary-based segmentation to reduce mis-segmentation of objects. Second, the process of object classification to separate bones and contrast-medium is carried out using a fuzzy classifier as knowledge based nonlinear classifier. Thereafter, either the CT numbers of pixels belonging to the regions segmented as contrast medium are substituted with their equivalent effective bone CT numbers based on segmented contrast correction (SCC) algorithm, or the classified regions as contrast medium can be converted to mu map using different calibration curves for energy mapping. The generated CT images were down-sampled followed by Gaussian smoothing to match the resolution of PET images. The bi-linear calibration curve was used to convert CT pixel values in HU to mu map at 511 keV. The visual assessment of segmented regions in clinical CT images by an experienced radiologist confirmed the accuracy of the algorithm for delineation of contrast enhanced regions. The results illustrate the difference between attenuation coefficients in the generated attenuation maps before and after SCC. Quantitative analysis of the generated mu maps from a clinical study showed an overestimation of 23.4% of attenuation coefficients in the 3D regions classified as contrast-medium. The algorithm is being refined and further validated in clinical setting to enable the application of this algorithm in PET/CT clinical arena.