Assessment of aortic valve tract dynamics using automatic tracking of 3D transesophageal echocardiographic images

Abstract

The assessment of aortic valve (AV) morphology is paramount for planning transcatheter AV implantation (TAVI). Nowadays, pre-TAVI sizing is routinely performed at one cardiac phase only, usually at mid-systole. Nonetheless, the AV is a dynamic structure that undergoes changes in size and shape throughout the cardiac cycle, which may be relevant for prosthesis selection. Thus, the aim of this study was to present and evaluate a novel software tool enabling the automatic sizing of the AV dynamically in three-dimensional (3D) transesophageal echocardiography (TEE) images. Forty-two patients who underwent preoperative 3D-TEE images were retrospectively analyzed using the software. Dynamic measurements were automatically extracted at four levels, including the aortic annulus. These measures were used to assess the software’s ability to accurately and reproducibly quantify the conformational changes of the aortic root and were validated against automated sizing measurements independently extracted at distinct time points. The software extracted physiological dynamic measurements in less than 2 min, that were shown to be accurate (error 2.2 ± 26.3 mm² and 0.0 ± 2.53 mm for annular area and perimeter, respectively) and highly reproducible (0.85 ± 6.18 and 0.65 ± 7.90 mm² of intra- and interobserver variability, respectively, in annular area). Using the maximum or minimum measured values rather than mid-systolic ones for device sizing resulted in a potential change of recommended size in 7% and 60% of the cases, respectively. The presented software tool allows a fast, automatic and reproducible dynamic assessment of the AV morphology from 3D-TEE images, with the extracted measures influencing the device selection depending on the cardiac moment used to perform its sizing. This novel tool may thus ease and potentially increase the observer’s confidence during prosthesis’ size selection at the preoperative TAVI planning.

Appendices

Appendix A

To better understand the AoA dynamics (potentially relevant for TAVI prosthesis sizing), Fig. 6 presents the average area/perimeter throughout the cardiac cycle over all subjects, as well as the relative area/perimeter change over time using the value at the end-diastole phase (last frame; ED) as reference. The average (and SD) eccentricity index (EI) was also calculated [34], and its relative change with respect to the ED phase is also depicted. Note that an EI of 0 represents a perfect circle, with higher EI indicating an elliptical geometry. Moreover, Fig. 7 illustrates the distribution of the temporal phase (as a percentage of the cardiac cycle duration) in which the maximum and minimum area and perimeter values were detected for the entire study population. Note that both first and last moments of the cycle correspond to the ED phase, with the average MS time point defined by the observer (used in the segmentation stage and set as the starting point for the tracking) also shown. Note also that, in the study population, the end-systolic (ES) moment occurred at a median ~ 53% of the R–R interval.

Overall, both area and perimeter (Fig. 6a–e) were larger in systole and lower in diastole, which is in agreement with previous reports that used independent delineation-based analyses in multidetector row computed tomography or 3D-TEE images [9, 12, 27, 28]. These results demonstrate the adequacy of the proposed software in extracting physiological and clinically-relevant measurements. Moreover, from the computed eccentricity index (Fig. 6c, f), one observed a more circular annular shape during systole, being more oval in diastole, which is also in accordance to the known dynamic conformational changes of the AoA throughout the cardiac cycle [10, 14, 27].

Fig. 6

Average a area, b perimeter, and c eccentricity index (EI) values measured across the cardiac cycle at the aortic annulus (AoA). Relative change of d area, e perimeter, and f EI values with respect to the end-diastole (last frame; 100% R–R interval). The vertical bars indicate the standard deviation of the measures across the entire study population. To obtain a temporal correspondence between frames of all measurement curves, the first and last frames of all sequences were aligned, and the curves’ values linearly interpolated over time

Fig. 7

Histogram of the temporal phase (as %R–R) at which the a maximum and b minimum area, and c maximum and d minimum perimeter were measured at the aortic annulus (AoA). The red dashed line represents the average mid-systolic phase (as %R–R) used by the observer to start the analyses (i.e. segmentation stage) across the study population. All measurement curves were temporally aligned as described in Fig. 6

The fact that the exact phase of maximum and minimum area/perimeter (Fig. 7) varied across the study population also corroborates the relevance of extracting these measures from multiple time points for an accurate prosthesis selection during preoperative TAVI planning, rather than from a single frame. Indeed, the phase of maximum size occurred, in median, at 20% of the R–R interval (area: 18.8%R–R, IQR [14.8, 27.0]; perimeter: 19.3%R–R, IQR [14.8 31.0]), as reported in Blanke et al. [12], which means that early systolic measures are slightly larger than those obtained at the mid-systolic phase, i.e. the indicated measurement phase in current guidelines [35]. In its turn, the phase of minimum size was, in median, at 80% the R–R interval (area: 81.7%R–R, IQR [63.0 92.9]; perimeter: 80.0%R–R, IQR [65.2, 92.5]), which is slightly later than the reported in a previous study [12]. This is partly related to a few cases in which the minimum value was found to be the ED frame (the last one in the sequence), which may be associated to some inaccuracy of the tracking algorithm given its distance from the mid-systolic reference frame and thus cumulated error. This result might be improved if one minimizes the number of propagation steps, as reported in Queirós et al. [19]. However, this type of strategies can only be applied to sequences with a complete cycle.

For completeness sake, Fig. 8 presents the measurements’ variation across the cardiac cycle for the other three studied levels. Despite the similar trends in relative changes across the cycle for both area and perimeter (larger values in systole and smaller ones in diastole), a distinct magnitude of relative change was obtained over time for each level. Indeed, the LVOT and aortic annulus presented larger relative changes in area and perimeter across the cycle when compared to both SoV and STJ levels (Table 2), with the difference between maximum and minimum measured values being statistically different for all levels. This has been reported for the aortic annulus in several previous studies [9, 11,12,13, 30], although the exact amount of change or its significance varied between populations (e.g., with or without aortic stenosis) and between studies (as described in the recent review on the subject in Suchá et al. [10]). Similar findings have also been described for the LVOT [14, 31]. Of note, when splitting the study population on the basis of the calcification severity (none/mild vs. moderate/severe cases; Table 1), no significant differences were observed for the measures’ relative changes at any of the studied levels (Appendix B).

Fig. 8

Average a–c area, d–f perimeter, and g–i eccentricity index (EI) values measured across the cardiac cycle at the left ventricular outflow tract (LVOT), sinuses of Valsalva (SoV) and sinotubular junction (STJ), respectively. Relative change of j–l area, m–o perimeter, and p–r EI values with respect to the end-diastole (last frame, 100% R–R interval). The vertical bars indicate the standard deviation of the measures across the study population. To obtain a temporal correspondence between frames, the first and last frames of all sequences were aligned, and the curves’ values linearly interpolated over time

Interestingly, the observed relative changes were lower for perimeter, when compared to the area. Such observation has been previously reported for the aortic annulus [12, 13], and has led to the suggestion of using perimeter-based values to size the TAVI prosthesis [36]. Note that this lower relative change for perimeter is not only associated to the relative change in AoA size (i.e. mathematically, these two quantities vary in different proportions with respect to the radius), but is particularly associated to the change in the AoA shape itself (see the changes in EI across the cycle in Fig. 6c, f), making the perimeter more robust (i.e. less variable) to differences throughout the cardiac cycle.

With respect to the EI (Figs. 6c, f, 8g–i, p–r), the AV tract presented a more circular geometry at the SoV and STJ levels, when compared to either AoA or LVOT planes. Moreover, the relative change of EI was significantly larger at the LVOT level (in a paired t-test against the other levels), which is in agreement with previous studies [14, 29]. In contrast, both SoV and STJ had smaller relative changes in EI across the cycle (in paired t-tests against both LVOT and AoA levels). Their larger area/perimeter values in systole can be mostly associated to an increase in size due to higher systolic than diastolic pressures. Note that, between them, the STJ presented a tendency for a lower relative change in area/perimeter values (Table 2; Fig. 8), which might be explained by the lower elasticity of this anatomical ring [26].

Appendix B

Table 4 presents the average (and standard deviation) of the maximum/minimum area and perimeter measured during the cardiac cycle, and associated absolute and relative differences, for the four studied levels, when splitting the study population on the basis of the calcification severity (none/mild vs. moderate/severe). Note that the calcification score (none, mild, moderate or severe grade) was visually determined by an experienced cardiologist using a preoperative MDCT scan at the time of TAVI planning and was retrospectively gathered for this study. Overall, despite the slightly different maximum/minimum values for some levels (namely LVOT, SoV and STJ), no statistically significant difference was found between groups for the measures’ relative changes (in a two-tailed unpaired t-test, with Welch’s correction). This result suggests that, for our study population, no differences exist in terms of magnitude of conformational changes between patients with lower (none/mild group) or higher (moderate/severe) amount of calcium.

Table 4 Maximum/minimum dimensions, and associated absolute and relative changes, for the four studied levels when dividing the study population based on the calcification score into “none/mild” and “moderate/severe” groups