High-resolution computation of 2-D motion from sonogram

Motion analysis in sonogram is used in the heavy ion radiation therapy in cancer treatment in order to decide irradiation timing. Some of the authors have proposed a method by which motion of internal organs, which is sensitive to the expansion and contraction of lung, is computed as a regression process with respect to a respiratory signal. In this method the motion is obtained with low-resolution, since this method is based on the gradient method, and then a low-pass filter should be applied to the sonogram. In this paper, multiple image sequences with different resolution at each other are hierarchically generated by applying band-pass filters. The information of the motion obtained in lower resolution image is propagated to higher resolution image, i.e., the motion is computed recursively. In the lowest stage, conventional method with an information of respiratory signal is applied. The effectiveness of the proposed algorithm is confirmed through real image experiments.