Investigating the Convergence of Phase Analysis Indices by Gated SPECT MPI With Common Filters and Reconstruction Methods

Background: The present study aimed to investigate the agreement and convergence of left ventricular dyssynchrony parameters extracted from phase analysis using GSPECT images under different conditions of filtration and reconstruction.Methods: The study population included 120 consecutive patients with normal or abnormal GSPECT MPI. All patients underwent a 2-day rest and stress sestamibi GSPECT MPI. The GSPECT images were reconstructed and processed using reconstruction methods including filtered back projection (FBP) with Butterworth (cutoff =0.45, order =5) and Metz (cutoff =0.9, order =6) filters and ordered-subset expectation maximization (OSEM) (subset=4,16; iteration =8) with Gaussian filters. Phase analysis (PA) parameters were evaluated globally and regionally (anterior, inferior, septum, lateral, and apex) in patients with normal MPI and those with abnormal MPI.Results: According to intraclass correlation (ICC) analysis, there was a significant and robust convergence between the OSEM (4,8) and OSEM (16,8) reconstruction algorithms, both of which were with Gaussian filters (P<0.001). Furthermore, there was a significant and robust convergence between FBP and Butterworth (cutoff =0.45, order =5) and between FBP and Metz (cutoff =0.9, order =6) in measuring PA parameters (P<0.001).Conclusions: The findings indicated that PA parameter values obtained from GSPECT MPI data with the FBP and OSEM image reconstruction methods were strongly correlated. However, the values of normal patients were dependent on the reconstruction technique. Therefore, reconstruction methods should not be used interchangeably.