A new method for monitoring PET DQA results

Objectives The GE Healthcare PET systems acquire daily quality assurance (DQA) data using a rotating Ge-68 pin source. However, the default program does not alert the user to subtle changes in system performance and is unable to track results over time. A new program has been developed for analyzing PET DQA results over time in comparison to a reference scan.

Methods The new analysis program analyzes the coincidence counts from a DQA scan using a structural similarity image metric (SSIM). The SSIM is calculated for each block and is based on the mean coincidence counts in a block from daily scan data and from a reference scan. The reference scan is defined as DQA data that was acquired immediately post-system calibration. The DQA scan is based on a one minute acquisition and not on a prescribed number of counts. As such, changes in SSIM over time are a result of either reduced counts due to Ge-68 source decay or a change in system performance. Therefore, an expected SSIM that models the decay of the source is calculated and the bias from the expected SSIM is calculated for each block. The new program was written in MATLAB and it reads in DQA xml files that have been transferred from the scanner via FTP.

Results The SSIM for each block is graphed to show the mean counts in a daily scan in comparison with the expected mean counts, taking into consideration the decay of the source from the time in which the reference scan was acquired. The median, maximum and minimum bias of the SSIM as a function of the number of days post-system calibration for each DQA scan is also plotted. Action levels have been set for a bias in SSIM greater than 0.6%, at which point physics and engineering support are notified. This program found examples of a decrease in system performance in cases in which the default DQA analysis program did not.

Conclusions This new method of analysis is able to quantitatively indicate when blocks are not performing as expected. This program is more sensitive to changes in system performance, is effective in tracking PET scanner performance over time, and reduces the subjective evaluation of system stability.