Related Experiment Video
Updated: Jul 5, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
A fully automated deep learning approach for coronary artery segmentation and comprehensive characterization
Guido Nannini1, Simone Saitta1, Andrea Baggiano
1Department of Electronics Information and Bioengineering, Politecnico di Milano, Milan, Italy.
APL Bioengineering
|January 25, 2024
Summary
A new automated pipeline rapidly quantifies coronary artery calcium (CAC) and tortuosity (CorT) from CCTA scans. This tool reveals a negative correlation between vessel tortuosity and calcific plaque, aiding coronary artery disease risk assessment.
Area of Science:
- Cardiovascular Imaging
- Artificial Intelligence in Medicine
- Medical Image Analysis
Background:
- Coronary computed tomography angiography (CCTA) is crucial for assessing coronary artery disease (CAD) markers like coronary artery calcium (CAC) and tortuosity (CorT).
- Manual analysis of these markers is time-consuming and prone to bias.
- Objective and automated quantification methods are needed for efficient CAD risk stratification.
Purpose of the Study:
- To develop and validate a fully automated computational pipeline for segmenting coronary arteries and objectively quantifying CAC and CorT from CCTA images.
- To assess the correlation between coronary artery tortuosity and calcific plaque burden.
- To provide a tool for rapid and objective CAD risk assessment in clinical and population studies.
Main Methods:
- A supervised learning approach using a two-stage U-Net architecture (2.5D and 3D) was employed for coronary artery lumen segmentation.
- Geometric post-processing extracted vessel centerlines for tortuosity quantification.
- Image attenuation and region growing algorithms were used for automated CAC detection and scoring.
- The pipeline was trained and validated on 281 manually annotated CCTA datasets.
Main Results:
- The automated pipeline achieved high accuracy in coronary segmentation (Dice score: 0.896, Mean Surface Distance: 1.027 mm).
- Coronary artery tortuosity significantly increased from proximal to distal regions (p < 0.001).
- Conversely, calcium volume score was higher in proximal regions and showed a significant negative correlation with tortuosity score (p < 0.001).
- The complete analysis required less than 5 minutes per patient.
Conclusions:
- The developed automated pipeline provides a fast, objective, and accurate method for quantifying CAC and CorT from CCTA.
- A significant inverse relationship between coronary artery tortuosity and calcific plaque burden was identified.
- This tool can assist clinicians in CAD risk assessment by providing detailed morphological insights.
