Automated plaque characterization using deep learning on coronary intravascular optical coherence tomographic images

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Image Analysis

Background:

• Accurate coronary plaque identification is crucial for managing advanced atherosclerosis.
• Intravascular optical coherence tomography (OCT) provides detailed plaque morphology.
• Current plaque analysis methods can be labor-intensive and subjective.

Purpose of the Study:

• To develop and validate a fully-automated deep learning model for semantic segmentation of coronary plaque in OCT images.
• To assess the performance of the automated model in classifying lipidous and calcified plaque.
• To evaluate the accuracy of automated clinical lesion metrics derived from the model.

Main Methods:

• A deep learning model was trained and tested on a large, manually annotated clinical dataset of intravascular OCT images.
• Pixel-wise classification was used to identify lipidous and calcified plaque components.
• Performance was evaluated using sensitivity and specificity metrics.
• Automated lesion metrics were compared against ground-truth labels.

Main Results:

• The model achieved high sensitivities and specificities for pixel-wise plaque classification: 87.4%/89.5% for lipidous plaque and 85.1%/94.2% for calcified plaque.
• Automated clinical lesion metrics showed high agreement (<4% difference) with ground-truth labels.
• A-line classification derived from the model significantly outperformed previous deep learning approaches (p < 0.05).

Conclusions:

• Fully-automated semantic segmentation of coronary plaque using deep learning is feasible and accurate in intravascular OCT images.
• The developed model provides reliable plaque characterization and lesion metric quantification.
• This automated approach offers a promising tool for clinical decision-making and research in atherosclerosis.