Machine Learning Based 3D-Transesophageal Echocardiography Modeling for Transcutaneous Aortic Valve Replacement Compared to Valve sizing by Computerized Tomography

Background:

Prosthesis size selection for transcutaneus aortic valve replacement (TAVR) is commonly done by aortic annulus measurements in cardiac computed tomography angiography (CCTA). CCTA may be inadequate for patients with inability to coopearate, irregular heart rhythms or reduced renal function for accuracy and safetu reasons. Transesophageal echocardiographic 3D aortic root machine-learning based modelimg (3D-TEEm) may represent a valid alternative to CCTA based valve sizing.

We present our experience with 3D-TEEm valve sizing for TAVR in comparison to patients whose prosthesis sizing was done by CCTA.

Methods:

All consecutive patients that undewent 3D-TEEm sizing were included. Randomly age and gender matched patients that had conventional CCTA sizing, in 1:2 ratio were also included for comparison.

3D-TEEm sizing was done using Siemens eSie Valve (ver. 4.0). CCTA sizing was done using the Philps Intellispace portal (ver 8.0). Sizing parameters and accuracy as well as procedural variables and post procedural outcomes were compared.

Results:

25 patients assessed by, were matched to 49 patients assessed by CCTA. In the 3D-TEEm group, renal failure rate was significantly higher (table). Aortic annulus diameter, sinuses of valsalva diameter and prosthesis size was smaller in the 3D-TEEm group compared to the CCTA group. However prosthesis selection accuracy was similar. Fluoroscopically assessed paravalvular leak (PVL), amount of contrast medium and radiation doses were similar. Discharge echocardiography aortic gradients decreasd similarly, and PVL was comparable between groups. GFR remained unchanged compared to baseline.

Conclusions:

3D-TEEm of the aortic annulus for TAVR sizing seems a valid alternative to CCTA regarding accuracy of sizing , exposure to contrast or worsening renal function.