Impact of Motion Correction tools on reliability and agreement of three software packages for dynamic- perfusion PET/CT processing

Abstract

Abstract Funding Acknowledgements Type of funding sources: Public Institution(s). Main funding source(s): University of Groningen/University Medical Centre Groningen. Background PET/CT plays an important role in the evaluation of patients with CAD. It allows for non-invasive assessment of coronary function, by means of software based estimations of regional and global myocardial blood flow (MBF) and coronary flow reserve (CFR). However, measurements can differ between software vendors. Also, motion artifacts can impact their reliability. Literature stands that motion can lead to differences up to 500%. In-software motion correction (ISMC) tools are available in some processing software packages, however they rely on different algorithms and their specific role is yet unknown. Recently, an innovative data driven motion correction (DDMC) algorithm was developed, that specifically tracks and corrects for motion prior the reconstruction process. This opens up the possibility of an standardized MC technique. Purpose This study aims to determine the impact of ISMC and DDMC on the reproducibility and agreement of MBF/CFR values between three software packages for dynamic PET processing. It was hypothesized that the use of MC techniques improves agreement/reproducibility in these parameters between vendors. Methods Twenty-three patients categorized as normal after a 13N-ammonia PET/CT examination acquired between July-December 2020 were retrospectively included. Inclusion criteria: no history of CAD, no perfusion defects, normal LV-function and no MACEs and/or relapse at follow up. These criteria were used to prevent possible interference to MBF/CFR values by concomitant CAD. Reconstructions were made without motion correction (NMC), with ISMC and with DDMC. These reconstructions were analyzed in three different software: Corridor 4DM [4DM], Cedars-Sinai QPET [QPET], and SyngoMBF [Syngo]. Clinical data, as well as regional and global values of MBF in rest and stress and CFR were acquired for the final analysis. Paired T-tests, Bland Altman plots and intraclass correlation coefficients were used for statistical analysis. Results Study population consisted of 15(65%) woman, all patients >60 yo, and 21(91%) of adenosine stress (9% regadenoson). The impact of the MC techniques varied between technique, software and region. Differences ranged from −0.12 (QPET DDMC in RCA stress-MBF) to +0.81 (Syngo ISMC in RCA CFR) ml/g/min. QPET proved to be more sensitive to MC, as significant changes in all regional and global stress-MBF and CFR were detected after applying any MC technique (p<0.001). Bland-Altman plots showed heterogeneous results after the use of MC, with some territories improving the Bias and LoA but the opposite in others. Finally, absolute agrement by ICC was predominantly better when using DDMC compared to NMC and ISMC. Only DDMC achieved an ICC >0.9 (excellent) in the RCA CFR. Conclusions The use of DDMC improves the reliability of regional and global stress MBF and CFR in all packages when compared to the use of ISMC or NMC, however, the impact of DDMC is heterogeneous between the packages.