Reduction of attenuation correction artifacts in PET-CT

The conventional approach to attenuation correction in PET-CT, is to derive an attenuation map from the reconstructed high resolution CT image. For that purpose, the CT image is resampled to a lower resolution, the CT-values are converted to attenuation coefficients at 511 keV, and the resulting image is projected to generate the attenuation correction sinograms. In the approach followed here, the PET attenuation image is computed from the raw CT data rather than from the clinical CT-image. The CT data are first reduced to PET resolution and rebinned into parallel hole geometry. This reduced data set can be reconstructed with iterative algorithms in reasonable processing times. Existing algorithms have been combined into a procedure that generates attenuation maps in which the artifacts due to metal implants, oral and intravenous contrasts agents have been reduced. The procedure combines the "projection completion" method with maximum-likelihood reconstruction to suppress metal artifacts. The first evaluation in patients reveals excellent performance. To suppress the influence of intravenous contrast, attenuation values higher than that of water were simply replaced with the attenuation of water. This method was evaluated in seven patient PET-CT studies, each of them consisting of two CT-scans and one PET-scan. One CT scan was acquired before, the other one after administration of the intravenous contrast agent. In these 7 patients, 18 volumes of interest were studied. With the conventional method, the mean difference in standardized uptake value (SUV) produced with the two CT-scans was 3.3%, with a maximum of 11.5%. With the alternative conversion method, this difference was reduced to 1.5%, with a maximum of 8.5%. The comparison of the two CTs also revealed differences due to tissue and patient motion. The SUVs seem to be more affected by this motion than by the IV-contrast.