Quantification and Reduction of Respiratory Induced Artifact in Attenuation Correction of PET Data using Respiration Averaged CT: A Simulation and Phantom Study

Abstract

Purpose: Respiratory-induced artifacts are dominant in Positron Emission Tomography/Computed Tomography (PET/CT) images. We investigated the impact of using the ACT data (respiration-averaged CT) in attenuation correction process. We evaluated the improvement in parameters such as maximum standardized uptake value ( ) and size in different respiratory traces for multiple lesion sizes in various locations of the thorax and abdomen.
Procedures: The attenuation in PET sinograms were corrected using end inhalation CT (EICT), end exhalation CT (EECT), and average CT (ACT) respectively. It should be noted that stationary PET images (without the respiratory motion) were reconstructed, and evaluated as the stationary truth. For the phantom study, a moving phantom was built mimicking the respiratory movement. The attenuation in uncorrected PET data was corrected using the three CT images mentioned above.
Results: Using EICT for attenuation correction, the respiration pattern with 35-millimeter diaphragm motion results in a %53 error in  estimation in comparison with the stationary truth for a 9-milimeter lesion in the liver. The use of ACT in attenuation correction can reduce such amount of error in  estimation up to %10 for this lesion. For the phantom study, using ACT for attenuation correction results in significant improvement in Signal to Noise Ratio (SNR) and contrast (p-value<0.05). Besides, better  was acquired for all the lesions.
Conclusion: The amount of respiratory induced errors in the quantified values of both  and the volume of the tumor depends on the location of the tumor, its diameter, the amplitude of the diaphragm motion, and the CT image we use for attenuation correction. Overall, ACT shows better results in comparison with the aforementioned techniques for attenuation correction of PET data in thorax region.