European Heart Journal

Discussion

The 3D anatomic complexity of the AVAp is poorly depicted by 2D imaging technologies. In fact, the universally accepted misconception of an ‘aortic annulus’ has evolved from a simplistic interpretation of those 2D modalities and does not apply to any true anatomical structure, due to the crown-shape insertion of the aortic leaflets into the LVOT (Figure 1). With the advent of 3D imaging technologies, we have the opportunity to more appropriately describe the aortic valvar complex. In the present study, we validate a simple MDCTA post-processing method that reliably measures real anatomic landmarks relevant to the AVAp.

Imaging modalities currently used to assess the aortic valve have clear and concise limitations. The most commonly used, 2D echocardiography, is operator dependent and limited by the angle of incision of the ultrasound array. Furthermore, shadowing caused by heavy calcification (frequently observed in aortic stenosis patients) is a significant limitation of this technology. Among the limitations of computed tomography, motion artefacts and non-adequate image acquisition and post-processing are probably the most relevant ones. In fact, prior studies using MDCTA to assess aortic root dimensions, distance from the ‘annulus’ to the coronary arteries and valve area^15–20 have been limited by the use of measurements obtained in standard 2D (axial, saggital, and coronal) or in multiple oblique planes. Therefore, these studies faced some of the limitations of measuring in two dimensions (Figure 4).

In addition, earlier MDCTA reports^15,17 did not analyse inter-rater variability and reliability of the measurements. It would be most prudent to validate the accuracy and reproducibility of the method of measurement prior to its application in the clinical setting. Following this validation, the method could be assessed in well-designed prospective studies. The current study, therefore, is the first of several steps and a good beginning in attempting to validate this MDCTA methodology for accurate and clinically important aortic apparatus characterization. Prior MDCTA protocols have not been subjected to such scrutiny and validation, but nonetheless have been used clinically.

We decided to analyse patients with aortic stenosis because severely degenerative valves are the most difficult to characterize using echocardiography and CTA due to the high prevalence of calcium, which introduces discrepancies in the measurements. A protocol that demonstrates good inter-rater variability in non-diseased valves may not be as good for measuring calcified ones, but if the inter-rater reliability is good for calcified valves, it is expected to be even better for non-diseased ones, the real borders of which are easier to identify. Furthermore, characterization of non-diseased valves has by now no more than diagnostic implications. On the other hand, characterization of severely stenotic valves has a more immediate therapeutic implication, as in the selection of device size for TAVI.

To accurately measure diameters along a vessel, measurements need to be performed in a perpendicular plane to the long axis of this vessel. Therefore, we used a curved multiplaner reconstruction technique to measure diameters since this technique allows for generation of strictly perpendicular planes to a centerline at each point along the length of the vessel under study, reducing angulation errors inherent with the techniques used in previous reports. We only measured clearly identifiable landmarks that correspond to actual anatomic structures (ALBAP, SV, STJ, AA), and for angle measurements we constructed a virtual reference plane from actual anatomic structures. ‘Diameter’ measurements for the SV are not straightforward, due to the clover-leaf shape of their cross-sections and to the different sizes of each of the sinuses (Figure 2G). In oblique reconstructions, the distance measured across the SV may correspond to the distance from commissure to commissure in one patient and to the distance from cusp to commissure in the next one, therefore introducing heterogeneity in the measurements. To standardize the observed values, our methodology takes into consideration all available distances between opposite points (passing through the centerline) at the level of the SV, and it integrates them into a single value, which is the average distance (‘diameter’). The ‘ALBAP’ may be interpreted as the evolution of the 2D derived ‘aortic annulus’ misconception. Nevertheless, this 3D concept encompasses the fact of the oval shape of the LVOT and the multiple axes (‘diameters’) that can be obtained from it. The ventriculo-arterial junction, so clearly identifiable in pathological specimens, could not be reliably identified using this imaging modality. Since a portion of this landmark is in continuity with the fibrous mitro-aortic structure, its radiological density is similar to that of the adjacent tissue, which prevents accurate identification of its exact location. Therefore diameters of this structure were not calculated.

We confirmed the high level of reproducibility of our post-processing methodology for measuring diameters, despite the handicap of performing the measurements in sub-optimal image-quality sets (non-gated studies, extended use of 16 MDCT, and analysis of systolic phases). A high degree of inter-rater reliability and absolute agreement was demonstrated among the three raters.

The distance from the origin of the coronary arteries to the basal attachment of the related leaflet may be relevant information for the new TAVI techniques.¹⁰ Absolute agreement of these variables was observed to be lower, indicating that these measurements are less reliable. This may be in part due to some of the studies being non-gated, creating significant motion artefacts. Furthermore, limitations in the software used may be another source of variability, since the incremental measurements along the centerline are not equidistant.

Finally, we evaluated the angulation between the LV and the aorta by measuring the angle of insertion of the aorta into the LV. Although clinical implications of this angulation in the setting of percutaneous aortic valve implantation are unknown, it has been suggested by experienced operators that it might influence the success of the implant. Characterizing the spatial relation of the aorta and the LV using a single angle is a simplistic approach to this anatomic interaction and relies on assumptions that may explain the lower reliability of the measurements. The aortic long axis is curved, and assimilating it to a straight line introduces a potential limitation of the technique. Furthermore, the lines of the LV inflow long axis and the aortic long axis will never cross because they lie in different planes, and therefore an imaginary 2D plane must be constructed to create an intersect (Figure 5). This requires building a plane that includes the inflow long axis of the LV and a second plane that includes the long axis of the aorta (Figure 5A–C), and these two planes must be rotated on their axes (Figure 5D) until they are parallel to each other. Once they are parallel, they can be projected onto a single imaginary plane (Figure 5E) were angle measurements can be made (Figure 5F). Current software limitations do not support such complex manipulation, and we defined a surrogated method for achieving this (Figure 3). The software module used allows accurate construction of the LV inflow long axis (Figure 3A, C, and D), but does not allow simultaneous construction of the centerline of the aorta. Therefore, angle measurements were not performed between these two centerlines. Instead of the centerline of the aorta, we used the line that followed the aorta in the plane in which the maximum visualization of the aortic root was achieved (Figure 3B and D) for angle measurements. We believe that future software should take into account the ideal method previously proposed for a more accurate measurement of the angle.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Two-dimensional approach to diameter measurements using CT. These images show the source of inter-reader variability using two-dimensional planes for diameter estimation. The red lines in A1, B1, and C1 represent the position of the corresponding coronal planes represented in A2, B2, and C2. As can be appreciated, the measured diameter will vary significantly depending on the position of the cutplane. This is a major limitation of echocardiography. CT post-processing methods previously reported partially overcome this handicap by simultaneously manipulating three perpendicular cutplanes where to centre the assessed structure.

We specifically investigated whether the degree of agreement among the three raters was different depending upon the processing level of the scanner and whether a study was gated by examining two groups representing the extremes of the data measurement continuum (non-gated/MDCT16 and gated/MDCT64). We believe that this data analysis served our purpose well from a statistical analysis viewpoint because it limited the number of multiple comparisons to 21 and thus, made the interpretation of results more manageable and reduced the potential or erroneous findings due to type I error. Despite these precautions, however, we must be mindful of the fact that these results are based on a substantially smaller subgroup sample size and they are, therefore, more likely to show greater variability than the total group results. The great majority of the ICCs was observed to be similar in both groups, as were the median differences on each of the seven measures obtained by all three raters. This indicates that we currently do not have enough evidence to suggest that the degree of agreement among our raters is different between the two groups. As we mentioned earlier, this may very well be the result of not having enough power to detect this difference due to the small subgroup size. Future research efforts may help elucidate the answer to this question.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Ideal measurement of the angulation between the aorta and the LV (assuming that the axis of the aortic root is a straight line). The long axis of the aorta and the inflow long axis of the left ventricle are represented by two lines that do not intersect. A stepwise approach to determining the angle between these two lines is represented here. (A) Two lines are built that represent the long axis of the aortic root (blue line) and the inflow long axis of the LV (red line). (B) A plane is constructed that rotates around the aortic long axis line (blue plane). (C) A second plane is constructed that rotates around the LV inflow long axis line (red plane). (D) These planes are rotated on their axes (curved arrows) until they are parallel to each other. (E) These two parallel planes are projected onto a single plane, wherein the angle between the LV inflow long axis and the aortic long axis may be measured (F).

The first of several steps towards the clinical validation of any diagnostic technique should always be, in our opinion, the validation of the post-processing/measuring protocols by demonstrating that they provide consistent and reliable measurements. Only when this has been proven should we climb the next steps of validation against established diagnostic modalities and clinical evaluation. Given the high reliability demonstrated by this protocol, we consider this analysis an important step in establishing standard criteria for the evaluation of the complex AVAp. Nevertheless, only well-designed prospective studies will determine if pre-procedural definition of the aortic anatomy and aorta–LV relationship are important determinants of percutaneous procedural success, and if these variables may be of use in patient and device selection in order to optimize the results.