A Gaussian-Based Guidewire Segmentation and Tracking Method for Teleoperated Robotic Intravascular Interventions

Abstract

Surgeons now rely on X-ray fluoroscopy to guide intravascular catheterization procedures during cardiovascular interventions. While this modality poses radiation exposure threats to the interventionists, use of teleoperated robot are being proposed to further reduce radiation exposure and as well provide a more ergonomic workflow. However, the real-time tracking and segmentation of intravascular tools, such as guidewire and catheter, within the blood vessel process is still challenging. In this paper, a Gaussian-based method is adapted for segmentation and tracking of guidewire in angiographic frames obtained during teleoperated robotic catheterization. By assuming the guidewire s curvature as a Gaussian process, Radon transformation is performed to extract features of guidewire pixels in the X-ray frames, and these were taken as input to the Gaussian process to segment the guidewire pixels from the background. Further, the prior segmentation information was used to predict the guidewire direction in the angiographs. Finally, performance of the segmentation results was evaluated with six existing methods using six binary pixel classification indicators. It was observed that the Gaussian-based method achieves good results in all the indicators, and surpasses the previous segmentation methods.