A retrospective evaluation of Bayesian-penalized likelihood reconstruction for [15O]H2O myocardial perfusion imaging

Abstract Background New Block-Sequential-Regularized-Expectation-Maximization (BSREM) image reconstruction technique has been introduced for clinical use mainly for oncologic use. Accurate and quantitative image reconstruction is essential in myocardial perfusion imaging with positron emission tomography (PET) as it utilizes absolute quantitation of myocardial blood flow (MBF). The aim of the study was to evaluate BSREM reconstruction for quantitation in patients with suspected coronary artery disease (CAD). Methods and Results We analyzed cardiac [15O]H2O PET studies of 177 patients evaluated for CAD. Differences between BSREM and Ordered-Subset-Expectation-Maximization with Time-Of-Flight (TOF) and Point-Spread-Function (PSF) modeling (OSEM-TOF-PSF) in terms of MBF, perfusable tissue fraction, and vascular volume fraction were measured. Classification of ischemia was assessed between the algorithms. OSEM-TOF-PSF and BSREM provided similar global stress MBF in patients with ischemia (1.84 ± 0.21 g⋅ml−1⋅min−1 vs 1.86 ± 0.21 g⋅ml−1⋅min−1) and no ischemia (3.26 ± 0.34 g⋅ml−1⋅min−1 vs 3.28 ± 0.34 g⋅ml−1⋅min−1). Global resting MBF was also similar (0.97 ± 0.12 g⋅ml−1⋅min−1 and 1.12 ± 0.06 g⋅ml−1⋅min−1). The largest mean relative difference in MBF values was 7%. Presence of myocardial ischemia was classified concordantly in 99% of patients using OSEM-TOF-PSF and BSREM reconstructions Conclusion OSEM-TOF-PSF and BSREM image reconstructions produce similar MBF values and diagnosis of myocardial ischemia in patients undergoing [15O]H2O PET due to suspected obstructive coronary artery disease