Materials and methods
Study population
11 subjects from the EurValve programme with suspected severe aortic stenosis on echocardiography were included [7]. All subjects had pre-TAVI/SAVR peak AVPG assessment with TTE, invasive cardiac catheterisation, and 4D flow CMR.
Ethics
Ethical approval was provided by the National Research Ethics Service, United Kingdom (17/LO/0283). Informed written consent was obtained from all subjects.
Non-CMR assessment of peak AVPG
Transthoracic echocardiography was performed according to the British Society of Echocardiography guidelines [8]. Severity grading of aortic stenosis was performed as per the European Society of Cardiology (ESC) consensus guidelines [5].
Invasive pressure gradients were obtained during cardiac catheterisation in all patients as part of routine care before TAVI/SAVR. Full details of the procedure to invasively assess peak AVPG are presented in Additional file 1: (item 1) and previously published reports [7]. Manual 4D flow CMR assessment was performed in MASS (version 2019 EXP, LUMC Netherlands). The semi-automated pipeline 4D flow CMR assessment was performed in CAAS (prototype version 5.2.1, Pie Medical Imaging, the Netherlands).
CMR protocol
CMR was performed on a 3.0 Tesla Phillips Healthcare Ingenia system equipped with a 28-channel coil and Phillips dStream digital broadband MR architecture technology.
The CMR protocol included baseline surveys, cines (vertical long-axis, horizontal long-axis, short-axis contiguous left-ventricle volume stack, 3-chamber, and aortic root). Cine images were acquired during end-expiratory breath-holds with a balanced steady-state free precession, single-slice breath-hold sequence. Procedures relating to 4D flow CMR pre-processing were delivered in accordance with established standards of practice [9]. Technical parameters for the CMR protocol are presented in Additional file 1: (items 2 and 3).
4D flow CMR assessment of AVPG
Post-processing and analysis were computed using CAAS MR Solutions (prototype version 5.2.1, Pie Medical Imaging, the Netherlands). Two assessors with at least two years of CMR experience independently calculated peak AVPG using the novel semi-automated pipeline method (CGC, PG). CGC repeated the analysis after four weeks, blinded to the results of the first analysis for intraobserver reproducibility analysis. When calculating peak AVPG, the assessors was blinded from the results of TTE and invasive assessments.
Two methods involving 4D flow CMR were used to determine peak AVPG. The manual method, as described in Archer et al. [7], and a novel semi-automated pipeline method. For both methods, aliasing correction was performed automatically in all three-phase directions. Any spatial misalignment with cine superimposition was manually corrected throughout the cardiac cycle before quantification.
The manual method involved plotting several multi-planar slices above the valvular plane to assess the quality of flow curves in the region of the vena contracta. The reformatted plane with the highest velocity and no artefact was then used to determine the peak pressure drop, calculated using the simplified Bernoulli Eq. (4Vmax2). This method is rigorously described within a previously published report [7].
The novel automated pipeline 4D flow CMR evaluation of AVPG involved three steps which are detailed in Fig. 1.
Statistical analysis
All continuous parameters are reported as mean ± standard deviation (SD). Correlation between TTE, 4D flow CMR (manual method), 4D flow CMR (semi-automated pipeline method) and the reference invasive pressure drop assessment were calculated using Pearson’s correlation coefficient. Agreement between the four methods was calculated using Bland–Altman statistics, where the mean difference between the two methods was reported as the relative risk of bias (measured in mmHg). The Shapiro–Wilk test assessed the normality of data. Since the data was normally distributed, parametric two-tailed paired samples t-test were used between peak AVPG assessment using invasive measures compared with TTE and the two 4D flow CMR methods. Interobserver and intraobserver variability were assessed using the interclass correlation coefficient (ICC). For all analyses, p < 0.05 was deemed to be statistically significant.