A method and an apparatus for fiducial-less tracking of a volume of interest (VOI) have been presented. In some embodiments, a pair of intra-operative images of a portion of a patient is generated during treatment of a target region in the patient to show a bony structure of the patient. The bony structure shown is movable responsive to respiration of the patient. Then the pair of intra-operative images is compared with a set of digitally reconstructed radiograph (DRR) pairs, generated from volumes of four-dimensional (4D) diagnostic imaging data, to determine a location of the movable bony structure that corresponds to a particular volume of the 4D diagnostic imaging data.