4D CT image reconstruction based on interpolated optical flow fields

In Computed tomography (CT) imaging, respiratory motions create artifacts that distort anatomic structures leading to serious difficulties in lung cancer diagnosis and treatment. 4D CT approaches try to remedy this problem by acquiring CT images synchronized to respiration. Reconstruction methods allow the correction of motions to create smooth images. However, such methods rely on the acquisition of a large number of images, increasing the risk of toxicity. In this paper, we develop a method to interpolate accurate CT images at any respiratory volume. The method consists of acquiring few images along with their corresponding respiratory volumes. The bidirectional registration of the two images bounding the target volume produces optical flow fields. The originality of our method lies in the volume-based interpolation of these vector fields at the target position. The weighted backward-mapping of the inverse of these fields produces the desired image. Experimentations using NCAT simulations and real chest images, acquired with RPM system, have been carried out. The results show the good potential of the proposed method.