Automatic Aortic Valve Cusps Segmentation from CT Images Based on the Cascading Multiple Deep Neural Networks

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiovascular Imaging

Background:

• Accurate morphological assessment of aortic valve cusps is crucial for effective treatment planning.
• Manual segmentation of aortic valve components from CT images is labor-intensive and time-consuming.

Purpose of the Study:

• To develop a fully automatic method for segmenting aortic valve cusps from CT images.
• To improve the efficiency and accuracy of obtaining morphological measurements for treatment planning.

Main Methods:

• A novel approach combining spatial configuration-Net and U-Net deep neural networks was employed.
• The method was trained and validated on 258 CT volumes, including cases with severe calcifications.
• Evaluation involved five-fold cross-validation to assess segmentation performance.

Main Results:

• The automated segmentation method achieved a mean processing time of 69.26 seconds for all CT volumes.
• High Dice Coefficients were obtained for the aortic root (0.95) and acceptable scores for individual cusps (0.67-0.70).
• Strong correlations were observed between automatically calculated measurements and manual annotations.

Conclusions:

• The proposed deep learning-based method enables automatic and efficient segmentation of aortic valve morphology from CT images.
• This automated approach can reliably provide measurement values essential for clinical decision-making and treatment planning.