Wavelet-based U-shape network for bioabsorbable vascular stents segmentation in IVOCT images

Area of Science:

• Medical Imaging
• Cardiovascular Engineering
• Artificial Intelligence in Medicine

Background:

• Coronary artery disease (CAD) is a major cause of death.
• Bioresorbable vascular scaffolds (BVSs) are used to treat CAD.
• Current segmentation of BVSs in Intravascular Optical Coherence Tomography (IVOCT) images is insufficient.

Purpose of the Study:

• To develop an advanced segmentation method for BVSs in IVOCT images.
• To improve the accuracy of differentiating stent struts from surrounding tissue.
• To enhance the clinical applicability of BVSs in CAD treatment.

Main Methods:

• A novel Wavelet-based U-shape network was designed.
• The network incorporates an Attention Gate (AG) and an Atrous Multi-scale Field Module (AMFM).
• A wavelet fusion module was used to integrate feature maps.

Main Results:

• The proposed model achieved high performance metrics: Dice (85.10%), Accuracy (99.77%), Sensitivity (86.93%), and IoU (73.81%).
• The AG, AMFM, and fusion module significantly improved detailed contextual information capture.
• The model outperformed existing BVS segmentation techniques.

Conclusions:

• The Wavelet-based U-shape network offers a significant advancement in BVS segmentation for IVOCT imaging.
• This improved segmentation holds promise for clinical applications in CAD management.
• The methodology may be adaptable to other complex medical image segmentation tasks.