Partial-Volume Effect in PET Tumor Imaging

Abstract

PET has the invaluable advantage of being intrinsically quantitative, enabling accurate measurements of tracer concentrations in vivo. In PET tumor imaging, indices characterizing tumor uptake, such as standardized uptake values, are becoming increasingly important, especially in the context of monitoring the response to therapy. However, when tracer uptake in small tumors is measured, large biases can be introduced by the partialvolume effect (PVE). The purposes of this article are to explain what PVE is and to describe its consequences in PET tumor imaging. The parameters on which PVE depends are reviewed. Actions that can be taken to reduce the errors attributable to PVE are described. Various PVE correction schemes are presented, and their applicability to PET tumor imaging is discussed.

Correlation of Primary Tumor FDG Uptake with Histopathologic Features of Advanced Gastric Cancer

“…The explanation for our result is that all primary tumors in the present study were >2.8 cm in size and the partial volume averaging effect would be minimal for primary tumor FDG uptake. It has generally been accepted that FDG uptake is underestimated because of the partial volume averaging effect of small-sized tumors <2 cm on PET/CT [44]. Kim et al [31] also reported a poor correlation between tumor size and FDG uptake in patients with advanced gastric cancer.…”

Adaptive threshold segmentation of pituitary adenomas from FDG PET images for radiosurgery

“…Intuitively, since the images show a marked difference between the normal and abnormal glycolytic functions, thresholding was one of the easier methods to adopt and implement, 22 , 23 , 24 , 25 provided it is optimized for the specific scanner. Since we were delineating small tumors, fixed threshold was not an ideal option, given its own limitations (26) . Since this data‐driven adaptive threshold method is based on the simple principle that the required threshold for segmentation is inversely proportional to T/B, it is easily understood by an average clinical user than achieving an intuitive understanding of the mathematical models for clinical implementation.…”

“…The low signal‐to‐noise ratio (SNR) increases the uncertainty in observed SUV value, which, in turn, increases the width of the Gaussian curve. Thus, the image of the container or tumor appears bigger than its actual size (26) . To obtain the true volume of the container or tumor during the segmentation process, a smaller volume compared to the visible volume of the image needs to be marked, which is done by using a higher threshold.…”

Partial Volume Correction Strategies in PET

“…The most sophisticated PVC strategies rely on an adjunct coregistered structural image in which MR imaging plays a pivotal role because of the better contrast between the gray and white matter compared with CT [7][8][9][10]. Now, however, the use of correlated anatomic information provided by the CT component of dual-modality imaging systems (eg, PET/CT) makes it possible to perform accurate PVC in other organs and tissues including cardiovascular [11], atherosclerotic [12], and whole-body oncologic imaging [13,14].…”

“…This method could be extended easily for use in cardiovascular PET imaging but would be challenging to apply in tumor imaging because of the difficulties in delineating accurately the metabolically active parts of a tumor (eg, in lesions with wholly or partially necrotic centers) using structural imaging [14].…”