Effective Methods to Correct Contrast Agent-Induced Errors in PET Quantification in Cardiac PET/CT

Abstract

In combined PET/CT studies, x-ray attenuation information from the CT scan is generally used for PET attenuation correction. Iodine-containing contrast agents may induce artifacts in the CT-generated attenuation map and lead to an erroneous radioactivity distribution on the corrected PET images. This study evaluated 2 methods of thresholding the CT data to correct these contrast agent-related artifacts. PET emission and attenuation data (acquired with and without a contrast agent) were simulated using a cardiac torso software phantom and were obtained from patients. Seven patients with known coronary artery disease underwent 2 electrocardiography-gated CT scans of the heart, the first without a contrast agent and the second with intravenous injection of an iodine-containing contrast agent. A 20-min PET scan (single bed position) covering the same axial range as the CT scans was then obtained 1 h after intravenous injection of (18)F-FDG. For both the simulated data and the patient data, the unenhanced and contrast-enhanced attenuation datasets were used for attenuation correction of the PET data. Additionally, 2 threshold methods (one requiring user interaction) aimed at compensating for the effect of the contrast agent were applied to the contrast-enhanced attenuation data before PET attenuation correction. All PET images were compared by quantitative analysis. Regional radioactivity values in the heart were overestimated when the contrast-enhanced data were used for attenuation correction. For patients, the mean decrease in the left ventricular wall was 23%. Use of either of the proposed compensation methods reduced the quantification error to less than 5%. The required time for postprocessing was minimal for the user-independent method. The use of contrast-enhanced CT images for attenuation correction in cardiac PET/CT significantly impairs PET quantification of tracer uptake. The proposed CT correction methods markedly reduced these artifacts; additionally, the user-independent method was time-efficient.