Automatic 3D Segmentation and Identification of Anomalous Aortic Origin of the Coronary Arteries Combining Multi-view 2D Convolutional Neural Networks

Ariel Fernando Pascaner ¹, Antonio Rosato ², Alice Fantazzini ³

⁰Department of Civil Engineering and Architecture, University of Pavia, Via Adolfo Ferrata 3, 27100, Pavia, Italy.

Journal of Imaging Informatics in Medicine +

|

February 12, 2024

View abstract on PubMed

Summary

This study introduces a deep learning method for automatically segmenting and classifying coronary artery anomalies from CT scans. This approach aids in faster diagnosis of anomalous origin from the aorta (AAOCA).

Area Of Science

Medical Imaging
Artificial Intelligence in Medicine
Cardiovascular Imaging

Background

Accurate segmentation and classification of coronary arteries are crucial for diagnosing conditions like anomalous origin from the aorta (AAOCA).
Manual analysis of computed tomography angiographies (CTAs) can be time-consuming and prone to inter-observer variability.

Purpose Of The Study

To develop and validate a deep learning-based automated method for segmenting the aortic root and coronary arteries.
To automatically classify coronary arteries as normal or exhibiting anomalous origin from the aorta (AAOCA) using CTA data.

Main Methods

Implementation of three single-view 2D Attention U-Nets with 3D view integration for segmentation.
Training on 124 CTAs including normal coronaries and AAOCA cases.
Utilizing a decision tree model for automatic classification of segmented geometries.

Main Results

Achieved median Dice score coefficients of 0.95 for the aortic root and 0.84 for coronary arteries on a test set (n=13).
Demonstrated excellent classification performance with 100% accuracy, precision, and recall for distinguishing normal coronaries from AAOCA.

Conclusions

A novel deep learning method effectively segments and classifies coronary artery origins from CTAs.
This automated approach shows potential for rapid screening and risk stratification of patients with AAOCA.

Keywords:

AAOCA Convolutional neural network Coronary arteries U-Net