Accuracy of the attenuation correction on small animal PET/CT

Abstract

Positron emission tomography (PET) is a functional medical imaging modality that is acquired after the administration of a radiotracer. PET imaging technique is based on the coincident detection of gamma photons of 511 keV. If any of the antiparallel gamma photons does not reach the PET detectors due to attenuation (scatter or absorption), a coincidence is not recorded, which means missing information. Attenuation is the largest correction that is applied to PET images in order to obtain an accurate quantification of radiotracer activity concentration. One approach to perform that correction is based on creating an attenuation map using a computed tomography (CT) image in which its Hounsfield units are transformed into attenuation coefficients at 511 keV by applying a bilinear approximation. Then, the attenuation map is used to correct PET data in the reconstruction process. There are no studies that compare non‐attenuation corrected PET images and attenuation corrected PET images acquired with the Argus PET/CT scanner with different energy windows and reconstruction methods from a practical point of view (quantification results). The aim of this study was to perform that comparison by means of three different experiments. Our results showed that attenuation correction has an impact on the image data and results are different depending on the Argus PET/CT reconstruction method and energy window used. For filtered back projection (FBP) and ordered –subset expectation maximization (OSEM) 2D reconstruction methods, image activity (counts per second) increases when applying the attenuation correction independently of the attenuation medium and the energy window. However, for OSEM 3D, the activity decreases. The absolute relative error between the estimated and real activity concentration either for non‐attenuation corrected PET images or attenuation corrected PET images was smaller than 5%. Finally, recovery coefficients for non‐attenuation corrected PET images are similar than the ones for attenuation corrected PET images. The segmentation rule does not affect the recovery coefficient calculation. However, when segmenting small objects, VOI mean value does not provide an accurate activity concentration.