Data driven gated PET: effects of acquisition time and misregistration with CT

Introduction: Data driven gating (DDG) can minimize issues with patient motion and enhance FDG-PET quantification. However, DDG utilizes <100% of PET data, leading to increased image noise unless acquisition time is increased. Misregistration between DDG-PET and CT may also occur, altering the potential benefits of gating. As the clinical usage of DDG-PET continues to grow, understanding how key parameters impact its utility becomes particularly relevant. In this work, the effects of PET acquisition time and misregistration with CT were assessed with a new DDG-PET/CT technique for correcting respiratory motion.

Methods: A GE D690 was used for FDG-PET/CT with 3 min acquisition times for most PET beds and a free-breathing helical CT. An extended acquisition time of 12 min was used in the primary PET bed where lesions of interest were located and respiratory motion effects were likely. A low-dose cine CT was also acquired over the same primary PET bed region, also with free-breathing. GE’s DDG PET method based on principal component analysis (Q.Static) was used with end-expiration data derived from ~50% of PET data at 30% from end-inspiration. For DDG-CT, end-expiration CT data was extracted from the cine CT by lung Hounsfield unit values and body contour using an algorithm developed in-house. Changes in PET acquisition time were simulated with retrospective reconstructions using different amounts of the total PET data from 30 sec to 12 min. Both the helical CT and DDG-CT were used for attenuation correction (AC) of DDG-PET data. SUV, SNR, and CNR were compared for 45 lesions in the liver and lung from 27 cases.

Results: Increases in PET acquisition time relative to the clinical standard of 3 min did not produce SUV bias. Reduced acquisition time down to 45 sec (25% of 3 min) also yielded unbiased SUV values. Only the largest reduction down to 30 sec (17% of 3 min) established statistically significant, positive SUV bias. This was true for both non-gated (NG) PET (bias: +6±13%; p=0.0041) and DDG-PET (bias: +13±23%; p=0.0016). DDG-PET (AC=helical CT) increased lesion SUV by 15±20% (p<0.0001), while DDG-PET/CT (AC=DDG-CT) increased SUV further by an additional 15±29% (p=0.0007). For DDG-PET, lesion CNR was statistically equivalent to 3 min NG-PET at both 3 min and 6 min acquisition times – an increased acquisition time of 12 min was needed to produce a clear increase in CNR (28±48%; p=0.0022). On the other hand, DDG-PET/CT at 6 min showed significantly increased CNR by 39±46% (p<0.0001) relative to 3 min NG-PET.

Conclusions: The common application of DDG-PET where counts are reduced by ~50% did not lead to inaccurate or biased SUV – the observed changes in SUV resulted from gating. Improved registration from DDG-CT was equally as important as motion correction with DDG-PET for increasing SUV in DDG-PET/CT. Overall, SUV quantification was more strongly affected by misregistration with CT than changes in acquisition time. Lesion detectability could be significantly improved when DDG-PET used roughly equivalent counts to NG-PET, but only when combined with DDG-CT in DDG-PET/CT. DDG-PET at 50% counts can provide enhanced SUV without adding bias, but misregistration with CT must also be addressed in order to take full advantage of its benefits for clinical applications.

Acknowledgements: This research was supported in part by NIH grants R21-CA222749-01A1, R03-EB030280-01, R01HL157273-01, and a ROSI grant from the UT M.D. Anderson Cancer Center, Division of Radiation Oncology. This research was conducted at the M.D. Anderson Cancer Center for Advanced Biomedical Imaging in-part with equipment support from General Electric Healthcare.

• Download figure
• Open in new tab
• Download powerpoint