AI-assisted computed tomography analysis for pre-procedural planning prior to TAVI

Area of Science:

• Cardiovascular Imaging and Interventions
• Artificial Intelligence in Medicine
• Medical Device Software

Background:

• Current software for cardiac CT analysis prior to transcatheter aortic valve implantation (TAVI) requires manual input and cannot perform complete analysis independently.
• There is a need for automated, efficient, and accurate pre-procedural CT analysis tools to aid in TAVI planning.

Purpose of the Study:

• To evaluate the performance of a fully automated, deep learning-based algorithm for pre-procedural CT analysis in TAVI patients.
• To compare the accuracy and correlation of the AI algorithm's measurements against the current clinical standard software.

Main Methods:

• A retrospective analysis of pre-procedural CT datasets from 98 patients undergoing TAVI for severe aortic stenosis.
• Comparison of measurements (annulus diameter, perimeter, area, and distances to coronary arteries) obtained from a standard TAVI CT analysis software and a fully automated deep learning-based CT analysis platform.

Main Results:

• Excellent correlation (Intraclass Correlation Coefficients > 0.95) was observed for annulus diameter, perimeter, and area measurements between the AI and conventional methods, with low mean absolute errors.
• Good correlation was found for the distances from the annulus to the left and right coronary arteries, indicating reliable anatomical assessment.
• The AI-based analysis demonstrated high accuracy, with mean absolute percentage errors for annulus diameter, perimeter, and area being 2.6%, 2.9%, and 4.8%, respectively.

Conclusions:

• The deep learning-based CT analysis demonstrated good to excellent correlation with conventional assessment for key pre-procedural TAVI measurements.
• Fully automated AI-based CT analysis represents a valuable and potentially more efficient alternative for pre-procedural planning in TAVI.