A CT-based deep learning system for automatic assessment of aortic root morphology for TAVI planning

Area of Science:

• Medical Imaging and Artificial Intelligence
• Cardiovascular Surgery
• Computational Anatomy

Background:

• Accurate pre-procedural planning for transcatheter aortic valve implantation (TAVI) is crucial for minimizing complications.
• Current methods for evaluating aortic root (AR) anatomy rely on manual 3D computed tomography (CT) analysis, lacking standardized automation.
• Automated solutions are needed to streamline the geometric analysis of the AR for TAVI candidates.

Purpose of the Study:

• To develop and validate a fully automated pipeline for geometric analysis of the aortic root (AR) using deep learning.
• To segment the AR, identify aortic annulus and sinotubular junction (STJ) contours, and quantify morphological biomarkers for pre-TAVI planning.
• To assess the accuracy and efficiency of the automated method compared to manual measurements by clinical experts.

Main Methods:

• Two 3D U-Net convolutional neural networks (CNNs) were trained on 310 CT scans for AR segmentation (Model 1) and contour identification (Model 2).
• The trained models were integrated into a fully automated pipeline for AR geometric analysis.
• The automated pipeline's results were validated against manual measurements from 178 TAVI candidates.

Main Results:

• The CNNs achieved high segmentation accuracy, with mean Dice scores of 0.93 for the AR and mean surface distances of 0.73 mm (annulus) and 0.99 mm (STJ).
• Automatic measurements showed good agreement with manual annotations, with minimal bias for annulus, STJ, and sinuses diameters.
• The automated method demonstrated effectiveness and speed in assessing AR anatomy, comparable to expert manual analysis.

Conclusions:

• The proposed fully automated pipeline provides an effective solution for quantifying morphological biomarkers essential for pre-TAVI planning.
• The AI-driven approach offers potential for significant time and cost savings in the pre-procedural assessment of TAVI candidates.
• This tool can enhance the accuracy and efficiency of anatomical evaluation of the aortic root in cardiovascular imaging.