Purpose: Carimas (Cardiac Image Analysis System) is a new software package developed at the Turku PET Centre for the quantitation of PET studies of the heart with a broad range of tracers. The goal of this study was to assess the reproducibility of results the package provides for myocardial perfusion (MP) quantitation using (15)O-labelled water.
Methods: Four observers with various levels of experience in nuclear medicine independently analysed 20 MP studies (10 rest flow: "rest", 10 adenosine-induced hyperaemia: "stress"). Each study was analysed twice. The linear mixed model for repeated measures was fitted to the data to calculate intraclass correlation coefficients (ICC), differences between the repeats (the intraobserver differences) and differences between the observers (the interobserver differences). Also, Pearson correlation coefficients (r) were calculated and Bland-Altman plots were drawn. The reproducibility of MP was assessed on global, regional and segmental levels. Thereafter, this analysis was applied in 48 consecutive clinical patients with suspected coronary heart disease (CHD).
Results: For the experienced observer the Pearson r for all segments was 0.974 at rest and 0.978 at stress (p < 0.0001), and the repeatability coefficients were 0.145 ml/g per min (15.5% of the average) and 0.389 ml/g per min (14.9%), correspondingly. The ICC reflected very good overall reproducibility. The intraobserver and interobserver differences were small, and the difference between the most and the least experienced observers at stress was 8.5% for the global MP. The clinical accuracy of the perfusion in the detection of CHD was excellent (positive predictive value 91% and negative predictive value 88%) against invasive angiography.
Conclusion: The results demonstrate high reproducibility of myocardial perfusion quantitation with (15)O-labelled water PET using Carimas. The results support the feasibility of robust analysis and good clinical accuracy.