Objective: Attenuation correction of PET data requires accurate determination of the attenuation map (mumap), which represents the spatial distribution of linear attenuation coefficients of different tissues at 511 keV. The presence of high-density metallic dental filling material in head and neck X-ray computed tomography (CT) scanning is known to generate streak artefacts in the resulting CT images and thus in the corresponding mumaps generated using CT-based attenuation correction. Consequently, an under/overestimation of activity concentration occurs in corresponding regions of the corrected PET images. The purpose of this study is to develop a simple yet practical approach for reduction of metallic dental implant artefacts in the generated mumaps.
Methods: Currently available sinogram-based metal artefact reduction (MAR) algorithms operate directly on the raw sinograms. These usually consist of huge files stored in proprietary format not easily disclosed by the manufacturers and thus are not straightforward to read and manipulate. The proposed method uses the concept of virtual sinograms produced by forward projection of CT images in Dicom format for MAR. The projection data affected by metallic objects are detected in the sinogram space through segmentation of metallic objects in the CT image followed by forward projection of the metal-only image. Thereafter, the affected sinogram bins are replaced by interpolated values from adjacent projections using the spline interpolation technique. The algorithm was assessed using a polyethylene phantom containing materials simulating different tissues and a dedicated jaw phantom scanned before and after the insertion of metallic objects, where the corrected and noncorrected mumaps were compared with the artefact-free mumap. In addition, the Jaszczak and standard germanium phantoms including four metallic inserts were scanned on a PET/CT scanner to evaluate the impact of the MAR procedure on PET data through the comparison of uncorrected and corrected PET images to the actual activity concentrations in the phantoms. The proposed algorithm was also applied to head and neck CT images of 10 patients with metallic dental implants.
Results: The MAR method proved to be practical in a clinical setting and reduced substantially the visible metal induced artefacts. The mean relative error in regions close to metallic objects is reduced by approximately 90%. The statistical analysis of the Jaszczak and solid Ge-68 phantoms PET images did not reveal statistically significant differences between the corrected and artefact-free images (P>0.05). Moreover, the evaluation of clinical studies did not reveal statistically significant differences between the attenuation coefficients of the corrected mumaps and the expected theoretical values.
Conclusion: The proposed MAR method provides a simple and convenient approach allowing correction for the presence of metal artefacts caused by dental implants without the need to manipulate the complex raw CT data. Further evaluation using a larger clinical PET/CT database is under way to evaluate the potential of the technique in a clinical setting.