Aims

The increasing numbers in transcatheter aortic valve implantation (TAVI) require a reliable and easy-to-perform computed tomography (CT) assessment for procedural planning. We retrospectively compared a fully automated CT analysis software for comprehensive annular assessment with a validated semi-automated approach, including its sizing prediction on procedural success.  

Methods

341 patients (82±7 years, 47% male) undergoing transfemoral TAVI with either a balloon-expandable (198 patients, Sapien 3, Edwards Lifesciences Inc.) or self-expandable prosthesis (143 patients, Evolut R, Medtronic Inc.) were analyzed. Prosthesis sizing for implantation was based on pre-procedural CT imaging using a validated semi-automated software approach (3men­sio Structural Heart, Pie Medical Imaging BV). Additionally, aortic annular and root dimensions were retrospectively analyzed using a fully automated software (Heart Navigator, Philips) and compared with semi-automated results. Based on these measurements, prosthesis sizing in relation to procedural outcome was evaluated.

Results

The automated software calculated a slightly, but significantly, larger mean annulus diameter (systolic: area-derived (AD) 24.7±2.0 vs. 24.1±2.0 mm; perimeter-derived (PD) 25.1±2.1 vs. 24.5±2.1 mm, diastolic: AD 24.3±2.1 vs. 23.3±2.1 mm, PD 24.8±2.1 vs. 23.8±2.1 mm, all p<0.001), a smaller annular plane-coronary ostium distance (right ostium 14.9±5.7 vs. 15.4±3.0 mm; left ostium 11.7±2.6 vs. 15.1±7.0 mm, both p<0.001) and predicted a more left anterior oblique and more cranial angulated orthogonal fluoroscopic projection (left anterior oblique 8.5±8.9° vs. 4.2±10.3°; cranial 3.4±9.0° vs. caudal 2.5±9.8°, both p<0.001) compared to the semi-automated approach. Pearson analysis revealed a very strong correlation of annular dimensions between both approaches (systolic: AD diameter r=0.879, PD diameter r=0.877, diastolic: AD diameter r=0.851, PD diameter r=0.850, all p<0.001). Though, in patients with severe annular calcification, correlation of annular dimensions declined (r=0.709, p=0.01). TAVI prosthesis size prediction agreed in 74% of patients (75% balloon-expandable; 73% self-expandable) between automated and semi-automated analysis. In the remaining cases, automated analysis suggested a larger prosthesis size in 24% of cases and a smaller size in 2%. In those patients, where the fully automated software suggested a larger prosthesis size than the implanted valve based on semi-automated analysis, a significant higher rate of early mild paravalvular regurgitation (55% vs. 38%, p=0.015) was found compared to those cases with congruent prosthesis size prediction. Trans-prosthesis gradient (p=0.430), rate of new permanent pacemaker implantations (p=0.106) or major peri-procedural complication rate (including all-cause death, cardiac death, myocardial infarction, stroke, p=0.501) did not differ significantly in patients where the fully automated software predicted a different prosthesis size than the semi-automated software.

Conclusions