Deep Learning-Based Fully Automated Aortic Valve Leaflets and Root Measurement From Computed Tomography Images　- A...

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Image Analysis

Background:

• Aortic root dilatation (RD) requires accurate anatomical assessment.
• Existing automated measurement algorithms need retraining for specific conditions.

Purpose of the Study:

• To retrain a deep learning algorithm for automated aortic valve/root measurements using CT data.
• To evaluate the clinical feasibility and accuracy of the retrained algorithm.

Main Methods:

• Retrained a deep learning algorithm using 67 ECG-gated cardiac CT scans from patients with RD.
• Evaluated the algorithm on 100 additional CT datasets (aortic stenosis, aortic regurgitation).
• Compared automated measurements with expert manual measurements and assessed measurement time.

Main Results:

• The algorithm provided 3D visualizations and automated measurements of the aortic valve/root.
• Moderate-to-high correlation was found between automated and manual measurements.
• Automated measurements were significantly faster (122s) than manual measurements (618-1126s).

Conclusions:

• The automated algorithm assists in evaluating aortic valve/root anatomy.
• It aids in planning surgical and transcatheter treatments.
• The algorithm saves time and minimizes workload for clinicians.