3-mensio software for the measurement of the aortic annulus before TAVI: agreement with multi-slice tomography and with echocardiography

Purpose: Preoperative cardiac imaging plays a central role in transcatheter aortic valve implantation. 3mensio software (3mensio Medical Imaging BV, The Netherlands) allows offline 3D reconstruction of the ascending aorta from DICOM images of multislice computed tomography (MSCT). Our aim was to assess the agreement among transthoracic echocardiography (TTE), MSCT and 3mensio for the measurement of the aortic annulus. Methods: Annulus diameter was measured using TTE and MSCT in 97 consecutive patients. The measurements with MSCT were performed by a radiologist both in the short axis view of the aortic root and in 3-chamber view; the average annulus diameter was calculated for each patient. In addition, an interventional cardiologist, blinded to MSCT results, measured the long-axis, short-axis and mean aortic annulus diameter with 3mensio at the level of the virtual basal ring (a plane including the basal attachment of the aortic leaflets). Results: Mean aortic annulus diameter was 22.3±1.9 mm by TTE, 23.4±1.6 mm by MSCT (p<0.0001 vs TTE) and 23.6±1.9 by 3mensio (p<0.0001 vs TTE; p=0.12 vs MSCT). Correlation between MSCT and TTE (r=0.773; p< 0.00001), 3mensio and TTE (r=0.703; p<0.00001), and 3mensio and MSCT (r=0.810; p<0.00001) was excellent. However, the Altman-Bland method showed a trend to overestimation of the annulus with both MSCT and 3mensio compared with TTE, with a mean difference between MSCT and TTE of 1.06 mm (limits of agreement -1.4 to 3.5 mm), and between 3mensio and TTE of 1.23 mm (limits of agreement -1.7 to 4.2 mm). The mean difference between MSCT and 3mensio was 0.17 mm (limits of agreement -2.0 to 2.4 mm). Agreement between MSCT and TTE was moderate (kappa = 0.59), with different Medtronic CoreValve sizing in 20 patients (20.6%). Agreement between 3mensio and TTE was also moderate (kappa = 0.61), with different CoreValve sizing in 19 patients (19.6%). Agreement between 3mensio and MSCT was good (kappa = 0.81), with different CoreValve sizing in 9 patients (9.3%). Patients with disagreement had a significantly more oval-shaped annulus at MSCT (p=0.05). Conclusions: Measurements of the aortic annulus by MSCT usually exceeded those of TTE of about 1 mm, although the 2 techniques were highly correlated. Measurements with 3mensio were very similar to those of MSCT, allowing the interventional cardiologist to obtain accurate aortic measurements from DICOM images. The agreement between TTE and both MSCT and 3mensio was moderate, the latter suggesting the choice of a larger prosthesis size in 20% of the patients. We are currently investigating the clinical implications of such discrepancy.