Evaluation of an automated fetal myocardial performance index

Abstract

Objectives: 1. To develop an automated system for calculating the fetal myocardial performance index (MPI), a non-invasive, pulsed-wave Doppler derived measure of global myocardial function, from raw ultrasound waveform data. 2. To compare automation with human evaluation of the fetal left MPI using an international team of fetal cardiology experts. Methods: A multicentre international study was undertaken using online technology to compare automated and manual MPIs. Images were used from a prior study of women with uncomplicated, morphologically normal singleton pregnancies at 16-38 weeks gestation. Individual Doppler ultrasound cardiac cycle images from the left ventricle of 25 cases were selected. Each waveform was used in triplicate and placed into a random order (n = 75). One senior observer blinded to this repetition at each of six centres manually calculated the MPI for the 75 waveforms online. The automated system was used to also measure the MPI for those waveforms. Intraobserver repeatability and interobserver reproducibility of MPI measurements were assessed using intraclass correlation coefficients (ICC) and 95% confidence intervals (95%CI). Bland-Altman plots, ICC and 95%CI were used to evaluate agreement between each observer's manual MPI measurements and corresponding automated measurements. Results: Automated measurements had superior repeatability (ICC = 1.00) compared to manual intraobserver repeatability (ICC ranging from 0.692-0.967) and manual interobserver reproducibility (ICC = 0.777). There was good agreement between automated and manual measurements of MPI (ICC 0.78-0.95). Conclusions: Automated fetal MPI provides superior repeatability and reproducibility to manual methodology. Universal application of this technique could facilitate translation of the MPI into a clinically useful tool. OC21.03 Beat-to-beat variability of the fetal myocardial performance index (MPI) Objectives: To determine whether there is beat-to-beat variability in the fetal left MPI as evaluated by an automated MPI measurement system and whether there is a correlation between the MPI and fetal heart rate (FHR). Methods: In this prospective cross-sectional study, women at 20-38 weeks gestation with uncomplicated, morphologically normal singleton pregnancies were recruited to undergo 2D Doppler ultrasound examinations. Multiple cineloops of waveforms suitable for measurement of the left MPI were acquired during a single examination in each fetus. Raw cineloop data was analysed by our automated MPI system (with established ICC of 1.0 for any given waveform) to produce a set of MPIs. The corresponding instantaneous FHR for each individual cycle that the MPI was calculated for was also measured. Results: Data from examinations of 29 fetuses was analysed, yielding a mean MPI of 0.52 and a mean FHR of 151. The median number of cycles examined per fetus was 70 (interquartile range: 31-115). There was marked beat-to-beat variability noted. When expressed as coefficient of variation, the median CoV was 8.7% (range 4.9% to 13....

Automation of the Fetal Right Myocardial Performance Index to Optimise Repeatability

“…This study continues our research group's previous work on the left MPI automation algorithm and extends the perfect reproducibility to the right MPI (ICC 1.00) [24], encouraging future clinical evaluation of automated MPI. The main limitation to manual MPI application is the lack of a universal reference range, in part relating to a lack of consensus on optimal methodology.…”

“…The reproducibility and feasibility of our left MPI automated algorithm has been evaluated and demonstrated a perfect ICC of 1.00, improving on the manual interobserver ICC of 0.78 [24]. There has been a focus on the left heart in the literature due to the ease of MPI acquisition in a single Doppler waveform.…”

Use of the Myocardial Performance Index in Decreased Fetal Movement Assessment: A Case-Control Study

“…Although we were unable to establish a clear clinical utility of the MPI, the frequency of DFM as a pregnancy complication and the lack of validated tools for DFM triage and risk assessment make further work on potential triage tools, such as the MPI, important. As recently found by our research group, use of an automated system for calculation of the MPI significantly increases the reliability and reproducibility of the MPI values and reduces potential measurement bias [21].…”

“…An automated MPI system, recently validated in a number of studies [21][22][23], was utilised to calculate both LMPI and RMPI. The algorithm was operated on the MATLAB environment (The MathWorks Inc., Natick, MA, USA), and blood flow velocity images in raw format were imported into the program for analysis.…”

Beat‐to‐beat variability of fetal myocardial performance index

“…We recently introduced a fully automated system for calculating fetal left MPI, demonstrating superior repeatability and reproducibility (intraclass correlation coefficient (ICC) of 1.0) compared with a manual methodology, eliminating entirely the subjectivity of manual measurement for any individual waveform. With the reproducibility and repeatability of automated MPI calculation thus confirmed, the automated MPI system provides an ideal platform with which to assess the stability of the fetal left MPI.…”