Author + information
- Philipp Blanke, MD,
- Eva Maria Spira, MD,
- Razvan Ionasec, PhD,
- Felix G. Meinel, MD,
- Ullrich Ebersberger, MD,
- Michael Scheuering, PhD,
- Christian Canstein, MSc,
- Thomas G. Flohr, PhD,
- Mathias Langer, MD and
- U. Joseph Schoepf, MD∗ ()
- ↵∗Heart and Vascular Center, Medical University of South Carolina, Ashley River Tower, 25 Courtenay Drive, Charleston, South Carolina 29425.
Computed tomography (CT) is increasingly used for prosthesis sizing in transcatheter aortic valve replacement (TAVR) because it enables 3-dimensional assessment of the complex aortic root anatomy, including the aortic annulus dimensions and the distance from the coronary artery orifices to the aortic annulus (1,2). CT can further predict appropriate C-arm angulation for orthogonal projection of the annulus plane. However, manual determination of these measurements is cumbersome and time-consuming. Alternatively, 3-dimensional cardiac CT datasets may be analyzed by automated computational algorithms (3). The aim of this investigation was to evaluate the accuracy and time-effectiveness of semiautomated model-based annulus computation compared with manual planimetry in TAVR patients.
Of 54 consecutive patients with severe symptomatic aortic stenosis and tricuspid valve anatomy undergoing dedicated electrocardiography-gated CT for TAVR planning, 4 patients were excluded due to insufficient image quality and 50 were included in this analysis (mean age 82.0 ± 5.8 years; mean aortic valve area 0.7 ± 0.2 cm2). CT data were reconstructed at 300 ms past the R peak (section thickness, 0.6 mm) and transferred to a dedicated post-processing workstation equipped with prototype analysis software (Heart Valve Analysis Protocol, Siemens Healthcare Sector, Forchheim, Germany) on the basis of a 3-dimensional anatomic model of the aortic valve (4) (Figs. 1A to 1C). Both manual and semiautomated assessments were performed independently by 2 observers with 5 years and no experience in interpreting TAVR-planning CT studies, respectively. The inexperienced observer was trained on 10 datasets before the study. Each workflow was repeated after a 4-week interval to define intraobserver variability.
For manual assessment, CT image data were reformatted to display the aortic annulus, defined by a plane transecting the basal attachment points of the aortic cusps. Planimetry of the aortic annulus was performed by manually tracking the luminal contours, yielding the cross-sectional area (A) and perimeter (P). The area-derived diameter and perimeter-derived diameter were calculated (DA = 2 × √(A/π) and DP = P/π, respectively). The distance from the aortic annulus plane to the lower edge of the coronary ostia was measured in a perpendicular fashion. Finally, the corresponding cranial/caudal angulation of the annulus plane for an orthogonal projection at a left anterior oblique angle of 10° was assessed.
For semiautomated assessment, datasets were automatically processed for identification of the anatomic landmarks. The aortic annulus contour was automatically delineated and then adjusted manually in all cases (Figs. 1D and 1E). The distances of the detected coronary ostia were displayed from the midpoint of the orifice to the annulus plane in an orthogonal fashion and then manually adjusted to the lower edge of the orifice (Fig. 1F). The corresponding cranial/caudal angulation of the annulus plane to achieve an orthogonal projection at a left anterior oblique angle of 10° was automatically displayed (Fig. 1G). The time required for the analysis including manual adjustments was recorded. Hypothetical valve sizing was on the basis of the DA for implantation of a balloon-expandable Edwards SAPIEN Heart Valve (Edwards Lifesciences LLC, Irvine, California) with the following incremental sizing regimen: 23-mm valve for a DA ≤22 mm, 26-mm valve for >22 mm to 25 mm, 29-mm valve for >25 mm to 28 mm.
The mean analysis time was significantly lower for the model-based measurements in both the experienced (26 ± 8 s vs. 98 ± 12 s, p < 0.001) and inexperienced (34 ± 11 s vs. 123 ± 18 s, p < 0.001) observers. All 3 basal hinge points were correctly identified by the semiautomated aortic valve model in 47 of 50 patients (94%), whereas in 3 patients (6%), 1 hinge point had to be manually corrected. Both coronary orifices were identified correctly by the model-based approach in all 50 patients.
For the experienced observer, there was no significant difference between the manual and model-based assessment of the DA and DP, coronary ostia distance, and angulation prediction. Bland-Altman analysis revealed no systematic bias (Figs. 1H and 1I). For the experienced observer, agreement for prosthesis sizing by both methods was found in 44 patients (88%, κ = 0.80). Similarly, there was no significant difference or systematic bias between both methods when assessment was performed by the inexperienced observer. Agreement of prosthesis sizing between both methods was found in only 36 patients (72%, κ = 0.54) for the inexperienced observer due to a greater variability in manual measurements. Agreement of prosthesis sizing between both observers was found in 45 patients (90%, κ = 0.82) for model-based measurements, but in only 40 patients for manual measurements (80%, κ = 0.80), indicating that the semiautomated approach may allow for a greater standardization of annulus measurements, particularly in less inexperienced observers.
Despite the availability of new automated systems, observers should still be proficient in the manual determination of all measurements required for TAVR planning. Furthermore, this study has the limitation that prosthesis sizing was on the basis of the valve model from a single vendor. Nevertheless, our data suggest that semiautomated morphological aortic annulus quantification enables fast and accurate procedural planning with excellent agreement in manual planimetry and has the potential to improve the workflow and standardize annular measurements in the evaluation of patients before TAVR.
Please note: Dr. Schoepf is a consultant for and/or receives research support from Bayer, Bracco, GE, Medrad Inc., and Siemens. Drs. Ionasec, Scheuering, Mr. Canstein, and Dr. Flohr are employees of Siemens. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- American College of Cardiology Foundation
- Binder R.K.,
- Webb J.G.,
- Willson A.B.,
- et al.
- Samim M.,
- Stella P.R.,
- Agostoni P.,
- et al.