Metal Artifact Reduction in CT-Based Attenuation Correction of PET Data Using the Virtual Sinogram Concept Mehrsima Abdoli, Mohammad Reza Ay, Alireza Ahmadian Dept. Medical Physics and Biomedical Engineering & Research Center for Science and Technology in Medicine & Research Institute for Nuclear Medicine, Tehran University of Medical Sciences, Tehran, Iran mabdoli@razi.tums.ac.ir , mohammadreza_ay@tums.ac.ir , ahmadian@sina.tums.ac.ir Habib Zaidi Division of Nuclear Medicine Geneva University Hospital Geneva, Switzerland habib.zaidi@hcuge.ch Abstract — The presence of high density materials in CT imaging is known to generate strong streak artifacts in CT images. This will obviously impact the generated μmaps, and as such could affect the reconstructed PET image during the CT-based attenuation correction (CTAC) procedure. Thus the attenuation maps (μmaps) generated from these images will likely propagate the artifacts to the resulting PET images. The main problem associated with current sinogram-based methods for metal artifact reduction is the difficulty in manipulating raw CT data which usually consist of large files saved in vendor’s proprietary formats. The proposed method directly computes a virtual sinogram through forward projection of CT images to overcome the above mentioned cumbersome inconvenience. The metallic objects in the CT image are first segmented. This is followed by forward projecting the obtained metal-only image. Then, missing projection data affected by metallic objects are detected and replaced by interpolated values from neighboring data using the spline interpolation functions. The algorithm is applied to a polyethylene phantom scanned before and after insertion of metallic objects. The corrected and non-corrected μmaps are compared to artifact-free μmap. It was shown that the mean relative error in regions close to metallic objects is reduced by 30% after applying our method. In another experiment a Jaszczak phantom is used to evaluate the results of the algorithm on reconstructed PET images. The activity concentration error produced in PET images is reduced by 85%. Moreover, the reconstruction of attenuation correction factors shows an obvious reduction of metal artifacts in the generated μmaps after applying the proposed algorithm. This study reports results obtained from a limited set of experimental measurements, further evaluation using clinical data sets is ongoing. Keywords- PET/CT; Attenuation Correction; Metal Artifact; Quantification; Spline Interpolation; Virtual Sinogram I. INTRODUCTION In recent years, positron emission tomography (PET) has become an effective method for clinical diagnosis and assessment of response to therapy in oncology. To accurately measure the activity concentration in different tissues of the body, attenuation correction (AC) of PET images is mandatory. One of the most widely used techniques for AC on combined PET/CT systems consists in using computed tomography (CT) images, owing to the fact that CT images contain information about attenuation properties of biological tissues. The advent of in-line PET/CT scanners further stimulated the clinical use of CTAC in addition to using CT images for anatomical mapping and localization of the abnormalities visible on PET images. However, the presence of high density metallic objects, such as dental fillings and hip prostheses, causes insufficient number of photons to reach the detectors. This results in the production of streak artifacts in CT images. These artifacts will likely propagate into PET images during the CT-based attenuation correction (CTAC) procedure and might deteriorate PET image quality and bias the quantitative analysis of radiopharmaceutical uptake. Therefore, before applying the CTAC procedure, metal artifacts must necessarily be removed. Several techniques have been proposed for metal artifact reduction (MAR). In general, these methods can be categorized in two groups namely sinogram based and image based methods. In the first group, the correction is implemented on acquired sinogram space. Linear interpolation of the missing data is one of these methods [1]. Since the difference between metallic objects and other tissues’ CT numbers is considerable, a simple thresholding can be used for segmenting the metallic objects. The extracted image is forward projected to determine the projections in the sinogram space which are affected by metallic objects. These projections are then replaced by linear interpolation of other projections in the same projection angle. Finally, the corrected image is obtained from the reconstruction of the corrected sinogram through inverse Radon transform. Cubic interpolation of unaffected projections is another way of replacing the missing projections [2]. Replacing missing projections by their unaffected correspondence is another method of this category [3]. In this method, instead of using an interpolation algorithm, missing projections are replaced by their corresponding unaffected projection, i.e. the opposite angular position in spiral scanning and the same angular position of the next slice in step scanning. The second group of metal artifact reduction methods is based on image rather than the sinogram. Iterative deblurring [4], knowledge based [5] and segmentation [6] techniques are some examples of image-based methods. These methods are obviously less accurate than those used in the first group, This work was supported by Tehran University of Medical Sciences and Research Center for Science and Technology in Medicine. HZ acknowledges the support of the Swiss National Science Foundation under grant No. 3152A0-102143.