Supervised segmentation for guiding peripheral revascularization with forward-viewing, robotically steered ultrasound...

Area of Science:

• Medical Imaging
• Interventional Cardiology
• Artificial Intelligence in Medicine

Background:

• Critical limb ischemia (CLI) affects 500,000 US patients annually, often requiring revascularization to prevent amputation.
• Current minimally invasive procedures for peripheral artery revascularization have a 25% failure rate due to chronic total occlusions.
• Improved guidewire navigation is crucial for successful limb salvage in CLI patients.

Purpose of the Study:

• To develop and validate an automated ultrasound image segmentation method for guidewire navigation in chronic occlusions.
• To enable direct visualization of guidewire advancement paths beyond occlusions using integrated ultrasound imaging.
• To enhance the success of robotic-assisted guidewire steering for peripheral revascularization.

Main Methods:

• A supervised U-net architecture was employed to segment B-mode ultrasound images from a forward-viewing, robotically-steerable guidewire system.
• 2500 simulated images trained the classifier to differentiate vessel walls and occlusions from viable paths.
• Performance was evaluated using simulated data, 3D-printed phantoms, and ex vivo porcine arteries, comparing against ground truth from microcomputed tomography.

Main Results:

• The U-net classifier achieved high performance (sensitivity 0.95, F1 score 0.96) compared to traditional methods.
• Accurate segmentation was achieved in artery phantoms with lumen diameters ≥ 0.75 mm (>90% accuracy).
• Ex vivo artery testing demonstrated average accuracy, F1 score, Jaccard index, and sensitivity exceeding 0.9.

Conclusions:

• Automated segmentation of ultrasound images for guidewire navigation in partially occluded peripheral arteries was successfully demonstrated using representation learning.
• This AI-driven approach offers a potentially fast and accurate method for guiding peripheral revascularization procedures.
• The technology holds promise for improving outcomes in CLI patients by enabling successful guidewire navigation through complex occlusions.