A simple test to determine the quality of your clinical PET images

Abstract

The objective of the study was to present a simple method for comparing clinical PET images to a set of increasing quality images. Those different quality images were obtained by varying the activity concentration and the acquisition time. Images of a Jaszczak phantom were acquired with scan times that were calculated with a spreadsheet application for a personal computer to obtain 500, 1000, 2000, 5000, 7000 and 9000 counts/4 mm(3) voxel. During a 10-h period, each scan was repeated with longer acquisition times to obtain the same number of counts in the reconstructed images, but with lower count rate. On the second day, the study was repeated, putting the phantom in a water bath to simulate larger patients. The quality of the images obtained with the phantom in water was worse than without, as expected. Phantom data demonstrated clearly the effect of higher counts on image quality. Good quality images were obtained with counts above 5000 counts/voxel. Patient data can be situated to the phantom image set by comparing the counts per voxel and the activity concentration. The counts per voxel in all the regions of interest on patient data, with the exception of the brain, were at sub-optimal level leading to decreased image quality. It is clear that better image quality can be achieved mainly by incrementing the scan time. Our PET system, however, allows doubling our standard injected activity to obtain more image counts without significant contrast loss. This simple test can be performed at any PET center to situate the quality of routine clinical PET images in comparison to the optimal possible for that system.