Year: 2010
Type: report
Abstract: Fluroscopic imaging provides means to asses the motion of the internal structures and therefore is of great use during surgery. In this report we propose a novel bottom-up approach for the segmentation of curvilinear structures in these images. The main challenge to be addressed is the lack of visual support due to the low SNR where traditional ridge-based methods fail. Our approach combines machine learning techniques, unsupervised clustering and linear programming / discrete optimization. In particular, numerous invariant to position/rotation classifiers are combined to detect candidates pixels of curvilinear structures. These candidates are grouped into consistent geometric segments through the use of a state-of-the art unsupervised clustering algorithm. The complete curvilinear structure is obtained through a linking of these segments using the elastica model in a linear programming framework in a first variant, and using discrete optimization in a second one. Very promising results were obtained on angioplasty guide wire segmentation in cardiac interventional images.
Source: HAL
Institution: General Electric (United States)
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2615069734
