A qualitative and quantitative assessment of the impact of three processing algorithms with halving of study count statistics in myocardial perfusion imaging: filtered backprojection, maximal likelihood expectation maximisation and ordered subset expectation maximisation with resolution recovery.

INTRODUCTION: Ordered subset expectation maximisation with depth-dependent resolution recovery (OSEM-RR) is a processing algorithm reported to improve images with halved tracer activity in myocardial perfusion scintigraphy (MPS) compared to filtered backprojection (FBP) using conventional activities. OSEM-RR has not yet been compared with maximal likelihood expectation maximisation (MLEM).
METHODS: 39 patients undergoing MPS and two anthropomorphic phantoms (one with, one without an inferior wall insert) had full-time (FT) and half-time (HT) SPECT datasets acquired simultaneously and processed by FBP, MLEM and OSEM-RR. Two experienced reporters scored images of all clinical studies (n=234) for conspicuity of a perfusion defect, with results being compared using Wilcoxon paired and Kappa tests. A quantitative assessment based on mean segmental pixel counts taken from numbers automatically displayed over the 20 segments of Cedars Sinai Autoquant QPS image were compared using Pearson's correlation and Bland Altman analysis.
RESULTS: A small but consistent superior concurrence between FT and HT datasets for OSEM-RR compared to FBP and MLEM was observed for both qualitative and quantitative analyses. OSEM-RR resulted in better definition of the inferior wall defect on the phantom study.
CONCLUSION: OSEM-RR appears superior to both FBP and MLEM in terms of handling reduced count statistics.