String matching techniques for high-level primitive formation in 2-D vascular imaging

Temporal matching in endoscopic images for remote-controlled robotic surgery

Temporal matching is applied in the frame of the formation of high-level entities in remote-controlled robotic surgery. The objective is to track tumor boundaries over time to improve the segmentation stage in each image of the sequence to facilitate the tracking and localization of the tumor. It makes use of an attributed string matching technique to find the correspondence between tumor boundaries over time. Relationships are then exploited to reconstitute the tumor boundaries and remove the inconsistencies coming from the detection errors. Input data are free form shapes of different length representing the tumor boundary, extracted at a previous stage.

Coronary extraction and stenosis quantification in X-ray angiographic imaging

This work describes a method to quantify stenosis in X-ray coronary angiography. Vascular edge extraction is first performed based on a deformable spline algorithm. It makes use of directional S-Gabor filters to build an external energy field that is then used in a snake optimisation scheme. A string matching technique is then applied to match the contour points and obtain a trace between the matched points. This trace allows then computing the vessel diameter and deriving quantitative stenosis measurements. Experimental results are presented on simulated data and real images.