A plaque recognition algorithm for coronary OCT images by Dense Atrous Convolution and attention mechanism

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Cardiovascular Imaging

Background:

• Manual plaque segmentation in coronary Optical Coherence Tomography (OCT) images is time-consuming and prone to inaccuracies.
• Existing automated segmentation techniques require further refinement for clinical application.

Purpose of the Study:

• To develop a novel deep learning algorithm for high-precision segmentation and classification of coronary artery plaques.
• To provide an efficient and accurate automated decision support tool for physicians.

Main Methods:

• A novel deep learning algorithm integrating Dense Atrous Convolution (DAC) and an attention mechanism was developed.
• A dataset of 760 original OCT images was augmented to 8,000 images for training and validation.

Main Results:

• The proposed algorithm achieved high dice coefficients: 0.913 for calcified, 0.900 for fibrous, and 0.879 for lipid plaques.
• Performance surpassed five conventional medical image segmentation networks.
• Demonstrated superior effectiveness in automatic coronary artery plaque segmentation.

Conclusions:

• The developed deep learning algorithm offers a significant advancement in automated coronary artery plaque segmentation using OCT.
• The algorithm provides a reliable and superior alternative to manual segmentation and existing automated methods.
• The enhanced dataset serves as a valuable resource for future research in cardiovascular plaque analysis.