Motion Detection in Fluoroscopic Image and its Application to Catheter Marker Tracking

Abstract

The preferred treatment for cerebral aneurysms is currently endovascular intervention. The real-time navigation of tools is conducted with fluoroscopic image guidance. This imaging technology is composed of single or biplane panels mounted on C- arms with multiple orientations capability. In theory, simultaneous biplane image acquisitions allow for a 3D reconstruction of the object of interest, in this case, interventional devices (guide wires, catheters, coils). These devices must be precisely located and tracked in real-time in the two fluoroscopic images for 3D reconstruction. However, the X-ray output is adjusted to reduce radiation exposure which results in low contrast and low resolution images. Thus, tracking algorithms require numerous enhancement treatments which are time consuming. Many approaches on tools segmentation and tracking have been proposed in literature with highly successful results, but no author has yet achieved a real-time process. This constraining objective is of utmost importance for effective clinical implementation of 3D road-map guidance. A new solution is proposed to combine and restrict the tracking algorithms to motion detected regions in the fluoroscopic image sequences. Motion regions are located with simple and fast fluoroscopic subtraction and digitized using threshold based on the standard deviation. This technique is tested on catheter markers tracking from in vivo fluoroscopic images acquired during cerebral embolization of aneurysms. The effective tracking of multiple markers was benchmarked at 3 to 5 frames per second. The motion areas restriction of algorithms can be implemented to other tools' tracking to eventually achieve a successful real-time 3D road-map guidance.