Artificial Intelligence-Based Algorithm for Stent Coverage Assessments

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Device Technology

Background:

• Neointimal hyperplasia after stent implantation is a key factor in stent thrombosis.
• Optical coherence tomography (OCT) is crucial for assessing stent strut coverage in vivo.
• Manual analysis of OCT images for neointimal coverage is time-consuming.

Purpose of the Study:

• To evaluate the preliminary performance of an AI-based tool for detecting and categorizing covered and uncovered stent struts.
• To assess the accuracy and efficiency of AI in analyzing OCT images for neointimal coverage.

Main Methods:

• Development of an AI algorithm using YOLO11 neural networks.
• Training and validation of the algorithm using OCT pullbacks.
• Comparison of AI performance against expert human analysts' consensus.

Main Results:

• The AI tool demonstrated satisfactory strut detection (92% PPV, 90% TPR).
• Higher accuracy was observed for covered struts (81% PPV, 85% TPR) compared to uncovered struts (73% PPV, 60% TPR).
• AI agreement with analysts was moderate (κ = 0.444).

Conclusions:

• The AI tool shows initial success in detecting stent struts, particularly covered ones.
• Classification of uncovered struts remains a challenge requiring further refinement.
• Larger clinical studies are necessary to enhance the AI tool's performance and clinical utility.