A Novel Approach for Reducing Dental Filling Artifact in CT-Based Attenuation Correction of PET Data

Reliable attenuation correction methods in PET require accurate determination of the attenuation map (μmap), which represents the spatial distribution of linear attenuation coefficients (LACs) at 511 keV for the region under study. Since CT image pixel intensities are directly related to the LAC of the corresponding tissue calculated from the effective CT energy, the μmap at 511 keV can be directly generated from the CT images. The presence of high density dental fillings material in head and neck CT imaging is known to generate strong streak artifacts in the μmap which will likely propagate to the resulting PET images during CT-based attenuation correction (CATC). The purpose of this work is to develop a fast approach for reduction of dental filling artifacts in the generated μmap. Currently available sinogram based metal artifact reduction (MAR) algorithms are based on correction of raw data sinograms which are huge files usually stored in proprietary format not generally disclosed by manufacturers and thus are not straightforward to handle. Our method uses the concept of virtual sinograms for implementation of MAR, which are produced by forward projection of CT images in Dicom format. The projection data affected by metallic objects are detected in the sinogram space by segmentation of metallic objects in the CT image followed by forward projection of the metal-only image. Thereafter the affected projections are replaced by interpolated values from adjacent projections using the spline interpolation technique. The algorithm was applied to a dedicated phantom experiment scanned before and after metal insertion, where the corrected and non-corrected μmaps were compared to the artifact-free μmap. It was observed that by using this fast method, the mean relative error in regions close to metallic objects is ~35% without correction and decreases to ~5% after correction.