Automated Artificial Intelligence Mapping of Coronary Plaque Calcification: A Comparison with Manual Intravascular...

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Interventional Cardiology

Background:

• Intravascular imaging, such as optical coherence tomography (OCT), is crucial for improving outcomes during percutaneous coronary intervention (PCI).
• Accurate and rapid interpretation of OCT images is essential for effective PCI guidance.
• Automated analysis of OCT-derived coronary calcium is needed to enhance clinical workflow.

Purpose of the Study:

• To develop and validate a novel artificial intelligence (AI) software algorithm for automated assessment of coronary artery calcification using intravascular OCT images.
• To compare the performance of the AI algorithm against manual expert analysis of OCT-derived coronary calcium.
• To evaluate the AI model's ability to quantify clinically relevant calcified plaque characteristics.

Main Methods:

• A deep neural network (UNet-like architecture) was developed and trained on expert-annotated intravascular OCT pullbacks.
• The AI model was validated on independent datasets of previously unseen OCT images.
• Performance was evaluated using metrics such as Area Under the Curve (AUC), diagnostic accuracy, and correlation coefficients for calcium scoring.

Main Results:

• The AI model achieved high performance in identifying calcified plaque, with an AUC of 0.96 and diagnostic accuracy of 73.3% in internal validation.
• External validation demonstrated a diagnostic accuracy of 74.8% in identifying calcified plaques.
• The AI model showed strong agreement with expert assessment for the calculated OCT-calcium score (ρ = 0.84).

Conclusions:

• An automated AI software algorithm offers a rapid and efficient method for comprehensive coronary calcium mapping in OCT images.
• The AI tool shows significant potential for improving the detection and assessment of coronary calcium in clinical practice.
• Further development and refinement of the AI algorithm are expected to enhance its capabilities for guiding PCI procedures.