A kernel-based method for real-time markerless tumor tracking in fluoroscopic image sequence

This paper presents a kernel-based method for tracking tumor motion in fluoroscopic image sequence for real-time image-guided radiation therapy (IGRT). The proposed method employs a histogram-based feature space, which is robust against image deformation, to represent a tumor target in a given fluoroscopic image. The target tracking is then formulated by maximizing a Bhattacharyya coefficient that measures the similarity between the target model and target candidates in the subsequent image. The numerical solution for maximizing the Bhattacharyya coefficient is performed by a mean-shift algorithm. Experimental results demonstrate that the proposed method is superior to conventional template matching-based methods in terms of accuracy and computational cost.