Identification and Quantification of Cardiovascular Structures From CCTA: An End-to-End, Rapid, Pixel-Wise,...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiology

Background:

• Manual segmentation of cardiac structures from coronary computed tomography angiography (CCTA) is time-consuming.
• Developing automated solutions is crucial for efficient cardiac analysis.
• Deep learning offers a promising approach for automating complex medical image analysis tasks.

Purpose of the Study:

• To design and evaluate an end-to-end deep learning solution for cardiac segmentation and quantification.
• To assess the accuracy and agreement of the deep learning model against manual annotations.

Main Methods:

• A U-Net-inspired deep learning model was developed.
• The model was trained and validated on CCTA scans from 166 patients.
• Cardiac structures including left ventricular volume (LVV), right ventricular volume (RVV), left atrial volume (LAV), right atrial volume (RAV), and left ventricular myocardial mass (LVM) were segmented and quantified.

Main Results:

• The deep learning model achieved a high overall median Dice score of 0.9246.
• Specific median Dice scores for LVV, RVV, LAV, RAV, and LVM were 0.938, 0.927, 0.934, 0.915, and 0.920, respectively.
• Model predictions showed strong correlation and agreement with manual annotations for all measured cardiac parameters (r values ranging from 0.78 to 0.98).

Conclusions:

• The developed deep learning model rapidly and accurately segments and quantifies cardiac structures from CCTA images.
• The model demonstrates high pixel-level accuracy and excellent agreement with manual segmentation.
• This automated solution has significant potential for research and clinical applications in cardiology.