An Improved Correlation Model for Respiration Tracking in Robotic Radiosurgery Using Essential Skin Surface Motion

Abstract

Stereotactic radiotherapy robots provide large benefits due to their high accuracy and stability. Using external features to correlate tumor motions is important for tumor tracking, a key technology in robotic radiosurgery. However, the correlations between external features extracted through skin surface movements and tumor motions vary in different thoracic and abdominal regions. External features with low correlation coefficients result in the degeneration of the chosen correlation model, thereby yielding a low fitting accuracy. To solve this problem, this letter proposes a method to extract highly representative surface motion features called essential surface motion. This method is based on selecting moving surface areas that are correlated with tumor motion according to high Pearson's correlation coefficients. The surface area is segmented into several regions before measuring their corresponding correlations with internal tumor motion. In consideration of the multidimensional structures of essential motion features, an improved correlation model based on the minimum cost function is proposed. Experimental results prove that our improved correlation model using essential surface motions as external features has a lower average error and absolute error than those that use surface marker motions.