Automatic stent struts detection in optical coherence tomography based on a multiple attention convolutional model.

Area of Science:

• Medical Imaging
• Cardiovascular Interventions
• Artificial Intelligence in Medicine

Background:

• Intravascular optical coherence tomography (OCT) is crucial for assessing stent placement during percutaneous coronary interventions.
• Manual analysis of OCT images for stent strut labeling is time-consuming and resource-intensive due to the density and small size of struts.

Purpose of the Study:

• To develop and validate an automated method for detecting stent struts in OCT images.
• To improve the efficiency and accuracy of quantitative analysis in clinical OCT applications.

Main Methods:

• A novel multiple attention convolutional model was designed for automatic stent strut detection.
• The model incorporates multiple attention mechanisms for enhanced feature extraction and fusion.
• Multiple anchor frames were integrated to improve the precise detection of small stent struts.

Main Results:

• The model achieved high precision (0.9790) and recall (0.9541) on a dataset of 4625 training and 1156 testing OCT frames.
• Detection speed was rapid at 25.2 ms per image, outperforming mainstream convolutional models.
• The model demonstrated robust performance across diverse imaging conditions and enabled clear 3D stent reconstruction.

Conclusions:

• The proposed automated model significantly reduces manual analysis time and medical resource allocation.
• Enhanced detection efficiency of tiny and dense stent struts facilitates the clinical application of OCT quantitative analysis.
• This technology supports more effective guidance for percutaneous coronary interventions.