Author + information
- Received January 30, 2013
- Revision received May 3, 2013
- Accepted May 9, 2013
- Published online September 1, 2013.
- Yusuke Watanabe, MD,
- Marie-Claude Morice, MD∗ (, )
- Erik Bouvier, MD,
- Tora Leong, MD, PhD,
- Kentaro Hayashida, MD, PhD,
- Thierry Lefèvre, MD,
- Thomas Hovasse, MD,
- Mauro Romano, MD,
- Bernard Chevalier, MD,
- Patrick Donzeau-Gouge, MD,
- Arnaud Farge, MD,
- Bertrand Cormier, MD and
- Philippe Garot, MD
- ↵∗Reprint requests and correspondence:
Dr. Marie-Claude Morice, Institut Hospitalier Jacques Cartier, 6 avenue du Noyer Lambert, 91300 Massy, France.
Objectives This study sought to evaluate the accuracy, reproducibility, and predictive value for post-procedural aortic regurgitation (AR) of an automated multidetector computed tomography (MDCT) post-processing imaging software, 3mensio Valves (version 5.1.sp1, 3mensio Medical Imaging BV, the Netherlands), in the assessment of patients undergoing transcatheter aortic valve implantation (TAVI).
Background Accurate pre-operative aortic annulus measurements are crucial for patients undergoing TAVI.
Methods One hundred five patients undergoing MDCT screening before TAVI were evaluated. Aortic annular measurement was compared between automated 3mensio Valves software and manual data post-processing software on a dedicated workstation; we analyzed the discrimination value of annulus measurement for post-procedural AR in 44 recipients of a self-expanding valve.
Results The automated 3mensio Valves software showed good concordance with manual MDCT measurements as demonstrated by Bland-Altman analysis. The automated software provided equally good reproducibility as manual measurement, especially for measurement of aortic annulus area (intraobserver intraclass correlation coefficients 0.98 vs. 0.97, interobserver 0.98 vs. 0.95). In 44 patients after implantation of a self-expanding valve, the valve diameter/CT-measured geometric mean annulus diameter ratio by automated 3mensio Valves software showed moderate and better discrimination ability in predicting post-procedural AR compared with manual measurement (p = 0.12, area under the curve 0.77, 95% confidence interval: 0.63 to 0.91, area under the curve 0.68, 95% confidence interval: 0.50 to 0.86, respectively).
Conclusions The automated 3mensio Valves software demonstrated reliable, reproducible aortic annulus measurement and better predictive value for post-procedural AR, suggesting important clinical implications for pre-operative assessment of patients undergoing TAVI.
Transcatheter aortic valve implantation (TAVI) has emerged as a viable therapeutic option for patients with severe symptomatic aortic stenosis (AS) who are ineligible or high risk for conventional surgical aortic valve replacement (1–4). Although this technique has reached relative maturity, residual paravalvular aortic regurgitation (AR) occurs frequently after TAVI, and even mild paravalvular AR is associated with increased mortality (5). The main cause of paravalvular AR is valve undersizing due to underestimation of the annulus size. Accurate annulus measuring is crucial for the selection of the appropriate valve size and for minimizing residual paravalvular AR.
Multidetector computed tomography (MDCT) offers a 3-dimensional alternative for image reconstruction of the aortic annulus in a proven reproducible fashion. This imaging technique may facilitate valve sizing and has, therefore, the potential to improving patient outcomes (6–10).
Standardization of measurements may increase intraobserver and interobserver reproducibility and thus reduce the frequency of procedure-related complications. The novel automated MDCT imaging post-processing software, 3mensio Valves (version 5.1.sp1, Pie Medical Imaging BV, Maastricht, the Netherlands) was designed to provide an accurate automated pre-operative aortic annulus measurement with high reproducibility and less image post-processing time.
The aim of this study was to evaluate the accuracy and reproducibility of an automated MDCT imaging post-processing software in the assessment of patients with severe AS undergoing TAVI, and to assess the predictive value of aortic annulus measurements with automated MDCT imaging software for post-procedural AR.
Study population and design
Between June 2010 and November 2012, a total of 459 consecutive high-risk patients with symptomatic severe AS treated with TAVI at our institution were prospectively included in our TAVI database. Electrocardiogram-gated MDCT data were available in 343 patients of the study population. A group of 105 randomly selected patients who underwent MDCT screening before TAVI were enrolled in this study. Patients with severe symptomatic AS (valve area ≤1.0 cm2) were considered candidates for TAVI if they had a logistic EuroSCORE >20%, if surgery was deemed to be of excessive risk due to significant comorbidities, or if other risk factors not captured by these scoring systems (e.g., porcelain aorta) were present. The decision to proceed with TAVI was discussed by a dedicated heart team including experienced clinical and interventional cardiologists, cardiovascular surgeons, and anesthesiologists. All patients agreed to participate in the study, and written informed consent was obtained in all cases.
MDCT image acquisition
All examinations were performed using a Philips Brilliance 64-slice MDCT scanner (Philips Healthcare, Best, the Netherlands). Standard technical parameters were used: gantry rotation time 300 ms, axial coverage 40 mm (64 × 0.625 mm), 120-kV tube voltage, 850 to 900 mA intensity with our modulation, and temporal resolution 165 ms. Retrospective electrocardiogram gating was performed. Contrast enhancement was achieved with 50 to 80 ml of iomeprol 400 mg/ml (Iomeron, Bracco Imaging, Milan, Italy). To achieve optimal synchronization, a bolus tracking method was used in the descending aorta. Additional beta-blockade was not administered in any case, due to potential hemodynamic instability in severe AS.
Manual MDCT image analysis
All image data were transferred to an off-line post-processing dedicated workstation (EBW, Philips Healthcare). The mid-systolic phase of the cardiac cycle was selected (30% of the RR interval). The thickness of the reconstructed image was 0.67 mm. The datasets were reconstructed to achieve a double oblique transverse at the level of the virtual ring (aortic annulus) described previously (7,8,11). The annulus surface area was then manually traced, and the orthogonal long-axis annulus diameter (lDiam) and short-axis annulus diameter (sDiam) dimensions were measured. The CT-measured geometric mean annulus diameter (mDiam-CT) was derived as: mDiam-CT = 2 × √(annulus surface area/π). This value represents the average of all annulus diameters, according to a previously described method (12). In addition, the distance from the coronary ostia relative to the aortic valve annular plane was measured. This manual annulus assessment using the dedicated workstation was performed by 2 experienced cardiac CT observers in our center (E.B. and T.H.). One observer was highly experienced in MDCT valvular assessments (level of proficiency 3; E.B.), and the other observer had less, but sufficient, experience (level of proficiency 2; T.H.) according to the American College of Cardiology/American Heart Association statement on competency in cardiac CT imaging (13).
Automatic MDCT image analysis
The same MDCT images of the aortic root (reconstructed at 30% of the RR interval) were retrospectively analyzed with the automated 3mensio Valves version 5.1.sp1 (3mensio Pie Medical Imaging, Maastricht, the Netherlands). From the 3 multiplanar reformation planes and the 3-dimentional reconstruction, the aortic root was automatically segmented and a centerline across the aortic lumen was displayed (Fig. 1A). A perpendicular plane along the centerline provided a short-axis view of the aortic valve annulus. The centerline and the perpendicular plane can be manually adjusted and positioned immediately beneath the lowest insertion points of all 3 aortic cusps to obtain the most accurate measurements (Figs. 1B and 1C). After validation of the centerline and the perpendicular plane, the software automatically displays the short-axis view of the aortic root (Fig. 1D). In addition, 2 orthogonal curved multiplanar reformation views and the double oblique views along the centerline were displayed to measure the height of coronary ostia (Fig. 2). The annulus surface area was then manually traced, and the lDiam dimension, sDiam dimension, and the height of the coronary ostia (Fig. 3) were also measured for comparison with the measurements obtained using a manual data post-processing dedicated workstation (EBW, Philips Healthcare). This automated annulus assessment using 3mensio Valves software was performed by 2 less experienced interventional cardiology fellows (level of proficiency 1; Y.W., T.L.) who had previously received appropriate explanations from an experienced observer.
Inter- and intraobserver agreement
Retrieved from 27 randomly selected data files, aortic annulus diameters were re-measured by another observer to determine interobserver agreement and by the same observers subsequently to determine intraobserver agreement in both automated and manual analyses. All observers were blinded to previous measurements.
Vascular access and valve selection
Patients were selected for TAVI via the transfemoral approach or alternative approaches depending on the size, calcification, and tortuosity of the iliofemoral arterial access. The type of valve prosthesis was selected according to the diameter of the aortic annulus measured by manual MDCT post-processing software using the calculated mean diameter (12). The Edwards valve (Edwards Lifesciences, Irvine, California) was used in patients with an annular diameter of 18 to 24.5 mm, and the CoreValve (Medtronic, Santa Rosa, California) for diameters of 20 to 26.5 mm. The transsubclavian or transaortic approach was used as an alternative route in cases of unsuitable femoral arterial access in recipients of the CoreValve, and the transapical, transsubclavian, or transaortic route as the alternative to suboptimal femoral access with the Edwards valve. The same criteria for bioprosthesis sizing and selection were applied throughout the study period.
All procedures were performed by an experienced team according to our standard operating procedures, as previously described (14).
Assessment of AR by echocardiography
Pre-discharge transthoracic echocardiography was performed in all patients by experienced echocardiographers. Semiquantitative grading of paravalvular AR was performed. AR was graded as none, mild, moderate, or severe according to the Valve Academic Research Consortium criteria (15).
Comparison of automated and manual MDCT annular measurements for the prediction of post-procedural AR
In order to compare automated and manual MDCT annular measurements with respect to their ability to predict post-procedural AR, 44 of 105 patients undergoing CoreValve implantation were analyzed. Only patients implanted with a CoreValve were studied because recipients of the Edwards valve had a low incidence of post-procedural AR. Post-procedural AR ≥mild was used as an endpoint. The difference between the nominal valve size and annular dimensions was measured by the following methods and subsequently assessed for prediction of post-procedural AR:
1. valve diameter minus mDiam-CT;
2. valve diameter/mDiam-CT ratio;
3. valve diameter/lDiam-CT ratio;
4. nominal valve area minus MDCT annular area;
5. nominal valve area/MDCT annular area ratio.
The external area of a fully expanded (i.e., nominal) valve is 5.31 cm2 for a 26-mm valve, 6.61 cm2 for the 29-mm valve, and 7.54 cm2 for the 31-mm valve. We used these values for analysis; however, the self-expanding valve area and diameter in vivo is lower than the nominal prosthesis area and diameter because of the valve characteristics.
Definitions of clinical outcomes
The clinical outcomes of this study (annulus rupture, device success, 30-day mortality) were defined by the Valve Academic Research Consortium criteria (15).
Normality of distributions for continuous variables was tested using the Shapiro-Wilks test, and data were analyzed appropriately thereafter. Pearson correlations were used to compare the manual and the automated measurement of the aortic annulus diameter. Concordance between the manual and the automated measurements of the aortic annulus was evaluated using Bland-Altman analysis. In addition, intraobserver and interobserver agreement was evaluated for both techniques by calculating the intraclass correlation coefficients (ICCs), with excellent agreement defined as an ICC >0.8. Receiver-operating characteristic (ROC) curves were generated using post-TAVI paravalvular AR ≥ mild. Areas under the curve (AUC) were compared for derived measures. All statistical tests were 2-tailed, and statistical significance was defined as p < 0.05. The data were analyzed using PASW Statistics version 19.0 (SPSS, Chicago, Illinois) and Stata version 12 statistical software (StataCorp., College Station, Texas).
The clinical characteristics of the study population are shown in Table 1. We evaluated 105 patients (58 men; mean age 83.4 ± 6.7 years; mean logistic EuroSCORE 20.8 ± 12.1%) with severe AS undergoing TAVI. All patients had good-quality MDCT images suitable for manual and automated off-line quantification of the aortic root assessment.
Comparison between automated and manual MDCT assessment of the aortic annulus diameter
The mean values of the variables measured using automated 3mensio Valves software and manual assessment are presented in Table 2. There were good correlations in the aortic annulus measurements between the automated 3mensio Valves software and manual assessment. Similarly, Bland-Altman analysis showed good agreement between both methods without significant bias (Figs. 4A to 4C). The average bias of sDiam-CT, lDiam-CT, and mDiam-CT between automated and manual assessment were 0.27 mm (95% confidence interval [CI]: −2.25 to 2.80 mm), 1.30 mm (95% CI: −1.20 to 3.80 mm), and 1.15 mm (95% CI: −0.97 to 3.27 mm), respectively. The correlation and agreement of left coronary artery ostium height and right coronary artery ostium height were considered acceptable results (average bias 1.61 mm, 95% CI: −3.40 to 6.62 mm, average bias 3.45 mm, 95% CI: −0.58 to 7.47 mm, respectively) (Figs. 4D and 4E).
Interobserver and intraobserver reproducibility
Table 3 shows interobserver and intraobserver reproducibility. The ICC for the inter observer and intraobserver reproducibility was satisfactory for both automated and manual assessment. The inter observer and intraobserver reproducibility of the automated 3mensio Valves software was as good as manual measurement, especially for the measurement of the aortic annulus area (intraobserver ICC 0.98 vs. 0.97, interobserver ICC 0.98 vs. 0.95, respectively).
ROC curve analyses of the prediction of post-procedural AR
In order to compare automated and manual MDCT annular measurements with respect to their respective ability to predict post-procedural AR, 44 patients undergoing CoreValve implantation were analyzed. Procedural characteristics are shown in Table 4. Post-procedural AR ≥ mild was observed in 14 (31.8%) patients. For the valve diameter/mDiam-CT ratio, automated assessment showed moderate and better discrimination ability compared with manual measurement (comparison ROC curves p = 0.12, AUC: 0.77, 95% CI: 0.63 to 0.91, p < 0.01 for automated assessment, and AUC: 0.68, 95% CI: 0.50 to 0.86, p = 0.06 for manual measurement, respectively) (Table 5, Fig. 5B). For valve diameter minus mDiam-CT, valve diameter/lDiam-CT ratio, nominal valve area minus MDCT annular area, and nominal valve area/MDCT annular area ratio, automated assessment also showed moderate and better AUC values compared with manual measurement (Table 5, Figs. 5A, 5C to 5E). The valve diameter/mDiam-CT and nominal valve area/MDCT annular area ratios with automated assessment showed the highest discriminatory values for post-procedural AR.
This study demonstrated that the automated MDCT post-processing software (3mensio Valves) provides a reliable and reproducible aortic annulus measurement in patients undergoing TAVI, with a better predictive value for post-procedural AR. This novel software standardizes all images post-processing and encourages widespread use of MDCT assessment of aortic annulus before TAVI.
Accurate measurement of the aortic annulus is crucial for precise valve sizing in patients undergoing TAVI. The aortic annulus is a complex 3-dimensional structure, and the virtual ring has an oval shape formed by the junction of the nadirs of all aortic valve leaflets at the distal part of the left ventricular outflow tract (7,8,16). Accurate measurement of the oval-shaped annulus should be performed in the transverse plane and perpendicular to the aortic root axis. Two-dimensional transesophageal echocardiography (TEE) has been used to measure aortic annulus diameter and for valve sizing; however, 2-dimensional TEE annular measurement may underestimate the “true” annular size because the 2-dimensional technique only provides a sagittal view of the aortic annulus (7–10,17). Three-dimensional TEE may allow an analysis of aortic annulus measurement with the transverse plane and perpendicular to the aortic root axis, but the image acquisition using 3-dimensional TEE requires special skills to obtain reliable data, and reproducibility is relatively low (17). Three-dimensional MDCT constructs an image that is orthogonal to the root of the aorta immediately below the nadir of the aortic cusps, allowing for short-axis, long-axis, and annulus area measurement, and these measurements appear to be more reproducible across multiple readers (11). Previous studies have highlighted the advantages of 3-dimensional MDCT-guided measurement of the aortic annulus due to the technique’s ability to reconstruct the oval shape of the annulus based on isotropic high resolution and less inter- and intravariability (7–10). However, the implementation of these techniques in the routine workup of patients before TAVI depends on the availability and experience of the team.
The novel automated 3mensio Valves software was developed as a dedicated software program for specific analysis of the aortic root and provides an accurate automated measurement of the aortic annulus in patients undergoing TAVI by automatically constructing an orthogonal image of the aortic root. One previous study demonstrated that this automated MDCT post-processing software permitted accurate, highly reproducible, and repeatable measurement and required less image post-processing time compared with manual assessment (18).
In our study, automated annulus assessment using 3mensio Valves software was carried out by 2 less experienced interventional cardiology fellows (level of proficiency 1). Even though the automated measurement was performed by less experienced observers, this automated software provided accurate measurements and similar or better interobserver and intraobserver agreement compared with manual measurement by experienced observers. Owing to its reproducibility across readers of varying experience, the automated MDCT post-processing software has the potential to be particularly useful as TAVI becomes increasingly used in new centers. Even in experienced centers, this automated software should contribute to the improvement of the time/efficiency ratio for pre-TAVI analysis.
Previous studies showed MDCT annular assessments to be superior to 2-dimensional TEE assessment in reducing post-procedural AR and strongly predictive of significant post-procedural AR (19,20). This study demonstrated that automated MDCT annular assessment may have more accurate discrimination ability in predicting post-procedural AR compared with manual measurement. This can be explained by the fact that annular measurements by automated software were slightly greater than by manual measurement. All these features make this automated software a valuable tool for pre-TAVI analysis.
Our study reports the results observed in a prospective, single-center TAVI cohort of limited size. Patient treatment bias is inherent in nonrandomized observational studies, and may have affected the clinical outcomes.
This study is a head-to-head comparison of 2 different post-processing MDCT software programs, and it is not yet clear whether these measurements are gold standard. However, previous studies have demonstrated 3-dimensional MDCT measurement of the aortic annulus to be one of the most accurate measurements for the assessment of prediction ability for post-procedural AR. The prevalence of significant post-procedural AR was relatively low, which may have precluded a thorough assessment of the program’s ability to predict post-procedural AR. Only patients implanted with a CoreValve were studied because recipients of the Edwards valve had a low incidence of post-procedural AR. The main cause of post-procedural AR is valve undersizing due to underestimation of the annulus size. The presence of calcification at the annulus level and valve underexpansion may also cause post-procedural AR (10,21). Implantation of the valve in a position either too high or too low may also play a role in the occurrence of AR (22). In this study, the amount of aortic annular calcification and the position of implanted valve were not evaluated. Because there were no cases of annulus rupture following TAVI, assessment of valve oversizing due to overestimation of the aortic annulus was difficult to perform. Moreover, the respective impact of automated and manual measurements on valve size selection was not evaluated. Further studies of larger patient populations are required to confirm our results.
The automated MDCT post-processing software (3mensio Valves) provided reliable and reproducible aortic annulus measurements in patients undergoing TAVI as well as better predictive value for post-procedural AR compared with manual assessment. This software standardizes all MDCT images post-processing and should prompt widespread use of automated MDCT assessment of aortic annulus in patients undergoing TAVI.
The authors thank Ms. Catherine Dupic for her assistance in the preparation of the manuscript.
Drs. Hayashida and Romano are proctors for transfemoral-TAVI for Edwards Lifesciences. Dr. Romano is a proctor for transapical-TAVI for Edwards Lifesciences. Dr. Lefèvre is a proctor for transfemoral-TAVI for Edwards; has received minor fees from Boston Scientific and Direct Flow Medical; and is a consultant for Symetis and Direct Flow Medical. Dr. Chevalier is a consultant for Medtronic. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- aortic regurgitation
- aortic stenosis
- area under the curve
- confidence interval
- intraclass correlation coefficient
- multidetector computed tomography
- geometric mean annulus diameter measured by computed tomography
- long-axis annulus diameter measured by computed tomography
- short-axis annulus diameter measured by computed tomography
- receiver-operating characteristic
- transcatheter aortic valve implantation
- transesophageal echocardiography
- Received January 30, 2013.
- Revision received May 3, 2013.
- Accepted May 9, 2013.
- American College of Cardiology Foundation
- Webb J.,
- Cribier A.
- Moat N.E.,
- Ludman P.,
- de Belder M.A.,
- et al.
- Zahn R.,
- Gerckens U.,
- Grube E.,
- et al.
- Messika-Zeitoun D.,
- Serfaty J.M.,
- Brochet E.,
- et al.
- Delgado V.,
- Ng A.C.,
- van de Veire N.R.,
- et al.
- Leipsic J.,
- Gurvitch R.,
- Labounty T.M.,
- et al.
- Blanke P.,
- Siepe M.,
- Reinohl J.,
- et al.
- Budoff M.J.,
- Achenbach S.,
- Berman D.S.,
- et al.
- Hayashida K.,
- Lefevre T.,
- Chevalier B.,
- et al.
- Kappetein A.P.,
- Head S.J.,
- Genereux P.,
- et al.
- Piazza N.,
- de Jaegere P.,
- Schultz C.,
- Becker A.E.,
- Serruys P.W.,
- Anderson R.H.
- Altiok E.,
- Koos R.,
- Schroder J.,
- et al.
- Willson A.B.,
- Webb J.G.,
- Labounty T.M.,
- et al.
- Jilaihawi H.,
- Kashif M.,
- Fontana G.,
- et al.
- Nombela-Franco L.,
- Rodes-Cabau J.,
- DeLarochelliere R.,
- et al.