An adaptive method of tracking anatomical curves in X-ray sequences

Abstract

Tracking anatomical structures in X-Ray sequences has broad applications, such as motion compensation for dynamic 3D/2D model overlay during image guided interventions. Many anatomical structures are curve-like such as ribs and liver dome. To handle various types of anatomical curves, a generic and robust tracking framework is needed to track shapes of different anatomies in noisy X-ray images. In this paper, we present a novel tracking framework, which is based on adaptive measurements of structures’ shape, motion, and image intensity patterns. The framework does not need offline training to achieve robust tracking results. The framework also incorporates an online learning method to robustly adapt to anatomical structures of different shape and appearances. Experimental results on real-world clinical sequences confirm that the presented anatomical curve tracking method improves the tracking performance compared to a baseline performance.