Assessment of a Novel Software Tool in the Selection of Aortic Valve Prosthesis Size for Transcatheter Aortic Valve Replacement

Abstract: Background. Transcatheter aortic valve replacement (TAVR) can be complicated by significant paravalvular leak (PVL). Optimal selection of TAVR prosthesis size can minimize the risk of clinically significant PVL. The aim of this study was to assess the utility of a proprietary software package, HeartNavigator (Philips), in selecting TAVR prosthesis to minimize PVL. Methods. All consecutive TAVR patients were considered for inclusion. HeartNavigator assessment was compared to three conventional (average, area-based, and circumference-based) computed tomography (CT) scan measurements of annulus diameter. The primary endpoint was clinically important (≥2+) aortic insufficiency (AI). Results. Fifty-six patients were suitable for analysis. The incidence of clinically important AI was 25%. The overall predictive value was identical for HeartNavigator (80.4%) and the three conventional CT parameters. Each method correctly identified a majority of patients destined for AI. Although HeartNavigator accurately identified a numerically greater portion of patients with AI (64.3%), this was not significantly different from the conventional CT parameters. Conclusions. As compared to conventional CT-based measurements, HeartNavigator offers an accurate method for selecting TAVR prostheses, comparable to conventional CT measurements.

J INVASIVE CARDIOL 2014;26(7):328-332

Key words: paravalvular leak, TAVR prostheses

_____________________________________

Transcatheter aortic valve replacement (TAVR) is a proven and established therapy in the management of patients with severe aortic stenosis who are at high risk for conventional, open surgical aortic valve replacement.^1,2 TAVR, however, in its current first-generation status, bears certain limitations. Among these is paravalvular leak (PVL). PVL of greater than mild degree (1+) is associated with adverse long-term prognosis and occurs in a substantial minority of TAVR patients.²

Selecting the size of the planned TAVR prosthesis in an effort to ideally match it to a patient’s aortic annular dimensions is paramount in preventing PVL. Assessing the annular size has evolved to include transthoracic and transesophageal echocardiography³ and planimetric imaging (computed tomography [CT] scanning and magnetic resonance imaging [MRI]).^4,5 Furthermore, for the planimetric methods, it has been recognized that carefully selecting the plane for measuring annular dimensions to correspond to the actual plane of the aortic annulus is superior to conventional anatomic orthogonal planes.⁶ This is most often performed manually and is simultaneously labor intensive and subject to interobserver variations.

The HeartNavigator software tool (Philips) is a proprietary, commercially available software program intended to simplify and automate the process of assessing the annulus and optimizing TAVR prosthesis selection. As the derivation of the annular plane is automated, it requires substantially less operator time and input than traditional CT measurements of annular size. However, the accuracy of this tool has not been rigorously compared to conventional CT measurements in a series of patients undergoing TAVR, especially in terms of predicting the occurrence of post-TAVR aortic insufficiency (AI). The aim of the present study was to assess the relative utility of HeartNavigator in comparison to conventional CT annular measurements in predicting clinically important PVL.

Methods

Patients. All consecutive patients undergoing TAVR at Henry Ford Hospital in Detroit, Michigan were potentially eligible for inclusion. Patients were required to receive a single implanted Edwards Sapien valve (Edwards LifeSciences)within a native stenotic aortic valve for inclusion in the study. Thus, patients who received “valve-in-valve” procedures within a prior surgically implanted bioprosthetic valve and patients who received more than one valve during their TAVR were excluded.

The analyses were conducted both retrospectively and prospectively. Patients whose TAVRs preceded the initiation of this investigation were analyzed retrospectively. The retrospectively analyzed group in turn included two subsets of patients: those in whom HeartNavigator had been employed as part of the decision-making process, and those whose TAVRs preceded the acquisition of HeartNavigator. For the latter subset, if their CT image acquisition had been compatible with HeartNavigator, their images were imported and the assessments were conducted by two investigators without knowledge of the size of the implanted prosthesis or the resulting degree of PVL. For the prospective subcohort, the analyses were performed a priori, and were openly reviewed and discussed at our weekly multi-disciplinary conference and formed one element of the decision-making process.

Description of HeartNavigator. HeartNavigator is a proprietary software package of the Philips Corporation. Although we consulted Philips personnel during the conduct of this study, Philips had no role in the design, conduct, or interpretation of this study. Consultations were on technical issues only and Philips personnel were never informed that we were conducting this analysis. There was no material sponsorship of this study.

The requirements for HeartNavigator are that CT imaging be performed with iodinated contrast agent, with slices of <1.5 mm thickness, and gated to one of two points in the cardiac cycle. The software program automatically identifies the principal cardiovascular chambers (left ventricle and aorta), and places fiducial marks at the key anatomic points of the hinge points of three aortic valve leaflets, and thus are intended to represent the location of the aortic annulus. The aortic valve plane is defined as the plane that touches all three leaflets of the valve from beneath.⁷ 

The fiducial point of each of the three cusp hinge points is not represented as an infinitesimal, one-dimensional point, but is represented as a sphere (or in two-dimensional space, a circle) of 4 mm diameter. The human operator has the opportunity to manipulate the images in three dimensions, primarily to assist in selecting a deployment angle for the TAVR, but which has no direct bearing on the annular size assessment.

The software then allows the operator to select virtual valves of 23, 26, or 29 mm diameter, corresponding to the three available dimensions of the Edwards Sapien valve. These virtual valves are superimposed on the three fiducial points and allow both qualitative and quantitative assessment of valve fit (Figure 1).

Conventional CT measurements. The selection of the plane of aortic annulus analysis was performed by a single skilled and experienced technologist who has primary responsibility for the clinical analyses of aortic annuli for all TAVR patients at our institution.

Two independent observers performed all measurements while blinded to one another’s results. Assessment of annular diameter was based on three assessments: the average of the maximum and minimum diameters (“D-average”), the diameter calculated from the planimetered area A [D = 2*√A/π] (“D-area”), and the diameter calculated from the planimetered circumference C [D = C/π] (“D-circ”). The measurements of the two operators were compared and when they differed by >1.5 mm, the operators repeated their measurements. If the difference persisted to be >1.5 mm, the two observers remeasured the dimensions together and reached consensus. The 1.5 mm criterion was selected arbitrarily, based on the consideration that it represents 50% of the diameter difference between successive sizes of the Edwards Sapien valve and thus would represent a potential threshold for selecting one prosthesis size over another, whereas smaller differences would not likely lead to differential decisions.

Analysis. The degree of fit of the virtual valve for HeartNavigator analyses was expressed as a percentage of overlap of the virtual valve and three fiducial symbols. While the “virtual” valve was always positioned with careful attention to symmetry, perfect symmetry was not obtainable. Manual movements of the virtual valve in a digital tool comes in increments of a single pixel or voxel, which may have asymmetric impact on the three fiducial symbols. The location of the three cusps in the majority of cases is not perfectly equidistant on a hypothetical round annulus, as in reality, most aortic annuli are oblong rather than round.⁸ Thus, given this asymmetry, in positioning the virtual valve, the overlap of the virtual valve and the three fiducial points is also asymmetric. Thus, we measured the diameter overlap of each of the three symbols (maximum = 12 mm, ie, 3 times 4 mm) and expressed it as a total percentage. The overlap was measured manually on digitized, magnified images. One might assume that the incremental overlap from the 23 to the 26 to the 29 mm valve is precisely an added 4.5 mm (corresponding to a 1.5 mm increase in radial distance), but this was generally not the case given the issues of asymmetry that we have emphasized. 

Aortic insufficiency was assessed based on the immediate postimplantation aortogram. Aortography was uniformly performed with all instruments withdrawn from the newly implanted aortic valve, and in a projection angle selected to ideally project the valve prosthesis orthogonal to the plane of the valve apparatus (ie, “squared up” on the profile of the metal cage). Aortic insufficiency was graded 0-4+ according to angiographic criteria.⁹

For the conventional CT measurements of annulus diameter, each of the three measurements were analyzed by expressing them as the difference between the measured imaging variable and the nominal dimensions of the implanted prosthesis. 

For all imaging variables (which were continuous variables), analysis first involved graphing the observed degree of AI (a dichotomized nominal variable) against the imaging variable. Subsequently, the sensitivity (ie, correct identification of patients with significant AI) and specificity (correct identification of patients without significant AI) were calculated across the spectrum of observed values. For each of the four variables, the overall predictive accuracy was calculated across the spectrum of observed values. The optimal diagnostic threshold was determined according to two criteria: (1) the majority of both sets of patients (significant or no significant AI) were identified correctly; and (2) the overall predictive accuracy was maximized. For each of the variables, the proportion of AI versus no-AI patients above and below the threshold was calculated and comparison was made using Chi-square analysis. Chi-square analysis was also utilized for comparisons of diagnostic accuracy between HeartNavigator and each of the other methods.

Results

Of our first 102 TAVR patients, fifty-six qualified for this study. As summarized in Figure 2, the most common reasons for exclusion were technical issues that precluded HeartNavigator analysis. This was primarily due to CT scans that were imaged and/or processed with protocols incompatible with HeartNavigator (mostly applicable to patients treated prior to our acquisition of HeartNavigator), the deliberate omission of contrast due to renal insufficiency, or other technical issues with CT (usually misregistration artifacts adversely affecting the accuracy of performing conventional CT parameters). Three cases were excluded due to technical issues with HeartNavigator (would not properly “track” the relevant cardiovascular chambers). 

The demographic, clinical, and hemodynamic baseline criteria of the included patients are summarized in Table 1. Patients were elderly, mostly female, and functionally limited. Hemodynamic data were reflective of severe aortic stenosis. 

The incidence of clinically important post-TAVR angiographic AI (≥2+) was 25%. The scatter plot of HeartNavigator values for patients dichotomized according to the degree of AI is illustrated in Figure 3. The corresponding calculation of the optimal threshold is illustrated graphically in Figure 4. Across the spectrum of observed values, the sensitivity (correctly identifying AI), specificity (correctly identifying the absence of clinically important AI), and overall predictive accuracy are depicted. The optimal diagnostic criterion was a 15% overlap of the virtual valve with the fiducial marks. Similar analyses were performed for each of the three conventional CT variables and are summarized in Table 2. Each of the four approaches achieved an identical overall predictive accuracy of 80.4% (P=NS). 

The performance of HeartNavigator is graphically illustrated in Figure 5. Adopting the 15% overlap criterion correctly classified a substantial majority of both patient subgroups.

Finally, Figure 6 illustrates the ability of each of the four approaches to correctly identify patients with significant AI. Numerically, HeartNavigator identified a larger proportion (64.3%) of those individuals destined for significant AI than any of the three conventional CT parameters (57.1%, 50%, and 50%, respectively). However, none of the differences between HeartNavigator and any of the other three methods reached statistical significance, which is likely attributable to the relatively small absolute number of patients with significant AI (n = 14).

Discussion

Our study validated that the HeartNavigator software tool is useful in supporting decision making in TAVR prosthesis selection. When the observed degree of overlap of the virtual valve exceeded 15% of the fiducial cusp markers, the incidence of important AI was 12.2% versus 64.3% of patients with overlap <15%. The sensitivity, specificity, and overall predictive value of HeartNavigator-based analysis were similar to those of more conventional CT-based measurements. Ours is the first study to demonstrate that this novel software tool can be employed with a similar degree or accuracy as the other more conventional CT-based measurements.

HeartNavigator offers the advantages of the automation of a number of steps, and minimal operator manipulation. The automation of landmark selection and identification of the annular plane combine to provide simplicity of use.

Our observation that a substantial minority of patients with significant AI are not correctly identified by any of the CT analyses underscores that geometry is not the sole determinant of post-TAVR AI.¹⁰ Given that the implanted prosthesis is nominally circular, whereas the annulus is almost never circular, the co-compliance of the annulus and prosthesis is a very important determinant. Prior studies have demonstrated that immediately after TAVR, the annular shape is partly forced into a less asymmetric configuration as compared to before TAVR.¹¹ Late imaging indicates that the annulus is even closer to circular,¹² suggesting that a remodeling process could be underway. The extent of calcification has also been implicated as a factor in the degree of post-TAVR AI,¹³ although this has not been consistently identified in all studies that have examined this question.¹⁴ Other variables that can contribute to the degree of post-TAVR AI and that will not be predictable from pre-TAVR imaging studies are procedural elements, such as the volume of the delivery balloon, completeness of inflation during deployment, and anatomic positioning of the prosthesis.¹⁵ 

Study limitations. The nature of TAVR precludes definitive assessment of the utility of any of the sizing decision tools. Only one prosthesis is implanted and thus the question of whether a different-sized prosthesiswould have resulted in more or less PVL remains unanswerable. Whether a different size of the implanted valve would have better prevented PVL remains conjectural.

We relied on angiographic assessment of AI. Angiographic criteria for AI do not differentiate between the relative contributions of PVL and central AI. However, all prior imaging studies have focused on total AI. Furthermore, total AI, and not specifically PVL, has been implicated in clinical outcomes.² Finally, most of post-TAVR AI is due to PVL, and not central AI due to leaflet non-coaptation.

One countervailing consideration to PVL is annular or aortic rupture, a complication that has been reported to be related to oversizing of the prosthesis.¹⁰ The incidence of this event in our series was too low to permit meaningful analysis. 

Conclusions

Our data indicate that HeartNavigator is a reliable tool for selecting the prosthetic size in cases of planned TAVR in patients with native aortic valve stenosis. Its overall accuracy in predicting the presence or absence of clinically important AI was comparable to those of conventional, established methods of CT-based annular size calculations.     

References

1. Leon MB, Smith CR, Mack M, et al; PARTNER trial investigators. Transcatheter aortic-valve implantation for aortic stenosis in patients who cannot undergo surgery. N Engl J Med. 2010;363(17):1597-1607.
2. Abdel-Wahab M, Zahn R, Horack M, et al. Aortic regurgitation after transcatheter aortic valve implantation: incidence and early outcome. Results from the German transcatheter aortic valve interventions registry. Heart. 2011;97(11):899-906.
3. Shahgaldi K, da Silva C, Bäck M, Rück A, Manouras A, Sahlén A. Transesophageal echocardiography measurements of aortic annulus diameter using biplane mode in patients undergoing transcatheter aortic valve implantation. Cardiovasc Ultrasound. 2013;11:5.
4. Rixe J, Schuhbaeck A, Liebetrau C, et al. Multidetector computed tomography is equivalent to transesophageal echocardiography for the assessment of the aortic annulus before transcatheter aortic valve implantation. Eur Radiol. 2012;22(12):2662-2669.
5. Quail MA, Nordmeyer J, Schievano S, Reinthaler M, Mullen MJ, Taylor AM. Use of cardiovascular magnetic resonance imaging for TAVR assessment in patients with bioprosthetic aortic valves: comparison with computed tomography. Eur J Radiol. 2012;81(12):3912-3917.
6. Achenbach S, Delgado V, Hausleiter J, Schoenhagen P, Min JK, Leipsic JA. SCCT expert consensus document on computed tomography imaging before transcatheter aortic valve implantation (TAVI)/transcatheter aortic valve replacement (TAVR). J Cardiovasc Comput Tomogr. 2012;6(6):366-380. 
7. Waechter I, Kneser R, Korosoglou G, et al. Patient specific models for planning and guidance of minimally invasive aortic valve implantation. In: Jiang T, ed. Medical Image Computing and Computer Assisted Intervention (MCCAI). Oxford University Press, Oxford. 2010:526-533.
8. Tops LF, Wood DA, Delgado V, et al. Noninvasive evaluation of the aortic root with multislice computed tomography implications for transcatheteraortic valve replacement. JACC Cardiovasc Imaging. 2008;1(3):321-330.
9. Baim DS. Grossman’s Cardiac Catheterization, Angiography, and Interventions. 7th ed. Lippincott, Williams, & Wilkins. Philadelphia. 2006:655.
10. Jilaihawi H, Kashif M, Fontana G, et al. Cross-sectional computed tomographic assessment improves accuracy of aortic annular sizing for transcatheter aortic valve replacement and reduces the incidence of paravalvular aortic regurgitation. J Am Coll Cardiol. 2012;59(14):1275-1286.
11. Blanke P, Reinöhl J, Schlensak C, et al. Prosthesis oversizing in balloon-expandable transcatheter aortic valve implantation is associated with contained rupture of the aortic root. Circ Cardiovasc Interv. 2012;5(4):540-548.
12. Willson AB, Webb JG, Gurvitch R, et al. Structural integrity of balloon-expandable stents after transcatheter aortic valve replacement assessment by multidetector computed tomography. JACC Cardiovasc Interv. 2012;5(5):525-532.
13. Ewe SH, Ng AC, Schuijf JD, et al. Location and severity of aortic valve calcium and implications for aortic regurgitation after transcatheter aortic valve implantation. Am J Cardiol. 2011;108(10):1470-1477.
14. Staubach S, Franke J, Gerckens U, et al; German transcatheter aortic valve implantation-registry investigators. Impact of aortic valve calcification on the outcome of transcatheter aortic valve implantation: results from the prospective multicenter German TAVI registry. Catheter Cardiovascular Interv. 2013;81(2):348-355.
15. Takagi K, Latib A, Al-Lamee R, et al. Predictors of moderate-to-severe paravalvular aortic regurgitation immediately after CoreValve implantation and the impact of postdilatation. Catheter Cardiovasc Interv. 2011;78(3):432-443.

___________________________________________

From the Center for Structural Heart Disease, Cardiology Division, Henry Ford Hospital, Detroit, Michigan.

Disclosure: The authors have completed and returned the ICMJE Form for Disclosure of Potential Conflicts of Interest. The authors report no conflicts of interest regarding the content herein.

Manuscript submitted November 20, 2013, provisional acceptance given December 9, 2013, final version accepted January 29, 2014.

Address for Correspondence: Paul T. Vaitkus, MD, Cardiology Division (Mail Stop 111D), Bay Pines VA Medical Center, PO Box 5005, 10000 Bay Pines Blvd, Bay Pines, FL 33744. Email: Paul.vaitkus@va.gov