Guidewire Endpoint Detection Based on Pixel Adjacent Relation in Robot-assisted Cardiovascular Interventions

Area of Science:

• Medical Imaging
• Computer Vision
• Surgical Technology

Background:

• Accurate visualization and tracking of endovascular tools are critical for successful interventions.
• Detecting endpoints of thin, flexible tools like guidewires in X-ray images is challenging due to their appearance and size.
• Existing segmentation methods struggle with endovascular tool detection, limiting procedural performance.

Purpose of the Study:

• To develop and evaluate a novel neighborhood-based method for detecting guidewire endpoints in X-ray angiograms.
• To improve the accuracy and reliability of endovascular tool localization for enhanced surgical navigation.
• To provide a robust solution for a common challenge in endovascular interventions.

Main Methods:

• A neighborhood-based approach combining pixel-level segmentation with a post-segmentation step.
• Utilizing adjacency relationships of pixels and skeletonization to identify guidewire endpoint pixels.
• Evaluation on a proprietary dataset from in-vivo rabbit studies.

Main Results:

• Achieved high segmentation performance with 87.87% precision and 90.53% recall.
• Demonstrated low detection error with a mean pixel error of 2.26±0.14 pixels.
• Outperformed four state-of-the-art detection methods in comparative analysis.

Conclusions:

• The proposed neighborhood-based method offers superior performance for guidewire endpoint detection in X-ray angiograms.
• This technique shows potential for generalization to other surgical tools and computer vision tasks.
• The method can enhance tool tracking and visualization systems, particularly in robot-assisted intravascular surgery.