P06.05: Automation of the fetal right myocardial performance index (RMPI) to optimise repeatability

Abstract

To compare the repeatability of an automated fetal right myocardial performance index (MPI) algorithm with manual measurements and to evaluate the degree of absolute agreement between values. A secondary aim is to determine the impact of MPI observer experience on these two aspects. A prospective cross-sectional study. Spectral Doppler double waveform, right MPI images were acquired, using a Voluson E8 from 65 uncomplicated singleton pregnancies at 22–39 weeks gestation. Three clearest cardiac cycles were selected from each case, by an experienced operator. Exported raw format images were analysed first with a MATLAB automated MPI software and then independently by an experienced and an inexperienced observer. Intraclass correlation coefficients (ICCs) and 95% confidence intervals (95% CI) were used to evaluate manual and automated intra- and inter-observer repeatability. Absolute agreement between the manual and corresponding automated MPI values was evaluated using ICCs, 95% CI and Bland-Altman plots. Successful automation was performed on 63 cases (97%) showing repeatability ICCs of: 0.83 for manual intra-observer; 0.77 for manual inter-observer and 1.00 for automated repeatability. The degree of absolute agreement between manual and automated values was: inexperienced observer ICC 0.43 (95% CI 0.21–0.62); experienced observer ICC 0.76 (95% CI 0.63–0.85). Right MPI may be a potentially useful tool in the early detection of cardiac dysfunction in a range of pathologies. Automation is precise, sensitive, demonstrates a superior reproducibility over manual measurements and reduces the experience required for successful analysis. This may lend a greater clinical applicability to MPI in functional fetal echocardiography, and future studies to develop a universal reference range based on automation would be useful. We aim for this, along with our Left MPI algorithm, to be made available as an open-source software application to apply to images from any ultrasound machine.