Abstract
523
Objectives: Recently a new technique for data driven gating (DDG) of whole body PET data acquired in continuous bed motion (CBM) was introduced (1). The technique is based on first transforming the CBM list mode PET data into a time series of 500 ms spatially filtered time of flight volumes. The respiratory signal waveform (WF) is then obtained by spectral analysis of the time varying displacement of activity distributions in these volumes. In this work, our aim was to evaluate the performance of this DDG technique versus the Anzai device and determine its impact on SUV measurements.
Methods: 15 patients (Mean Age: 63, BMI: 25, injected activity: 324 MBq of F18-FDG, imaged post injection: 68 min) were imaged on a SIEMENS mCT PET/CT scanner in CBM mode. All patients were fitted with the Anzai device to record their breathing WF. Following PET acquisition, the corresponding DDG WF for each patient was derived using the new DDG technique. The two WF (Anzai and DDG) were then compared by calculating their correlation coefficients over 8 anatomical zones and averaged over all patients. The 8 zones were: lower extremities to center of bladder(LEX-COB), center of bladder to center of right kidney(COB-CRK), center of right kidney to liver dome(CRK-LVD), liver dome to aortic arch(LVD-AAR), aortic arch to right lung apex(AAR-RLA), right lung apex to upper extremities(RLA-UEX), Center of Bladder to Aortic Arch(COB-AAR), and the entire signal (ES). Additionally, the acquired PET data for each patient was reconstructed using 3 methods: 1) static whole body [SWB], 2) HDChest using Anzai WF [HDC_AZ], and 3) HDChest using DDG WF [HDC_DG]. For the HDChest image, a 35% duty cycle was utilized. All reconstructions were performed with 3 iterations 21 subsets, TOF, PSF, and no post filtration. For each reconstruction we measured the SUVmax and SUVpeak on 12 liver and lung lesions, 1 splenic lesion, 1 gastric lesion, and for 1 patients with no lesions, measurements were made on a single renal pyramid. We then calculated the percent changes in SUVmax and SUVpeak with respect to SWB for each region and reconstruction. Finally we calculated the SUVmax and SUVpeak ratios for each region when measured using HDC_AZ and HDC_DG.
Results: The average correlation coefficients for each region were: 0.15 (LEX-COB), 0.69 (COB-CRK), 0.82 (CRK-LVD), 0.75 (LVD-AAR), 0.46 (AAR-RLA), 0.07 (RLA-UEX), 0.75 (COB-AAR), and 0.55 (ES). The highest correlation coefficients occurred between the kidney and the liver dome, both of which are subject to respiratory motion, and typically have high physiologic F18-FDG uptake. Conversely, the AAR-RLA, has relatively low amounts of motion and uptake, and resulted in the lowest correlation out of the regions impacted by motion. Measurements of SUVmax, SUVpeak, and percent changes with respect to SWB are tabulated below. In comparison to SWB, all motion correction reconstructions increased SUVmax and SUVpeak as a result of decreased motion. However, on average, the DDG reconstructions were lower by about 10% in comparison to the Anzai reconstructions.
Conclusions: These preliminary results show that the DDG algorithm produces respiratory WFs and SUV measurements that are highly correlated with those from the Anzai device particularly in areas affected by respiratory motion and warrants an evaluation in a larger patient population.
1) Schleyer P, Hong I, Jones J, Hamill J, Panin V, and Fuerst S. Data-driven respiratory gating whole body PET using continuous bed motion. Conf. proceed. IEEE MIC 2018, Sydney, Australia