Three-dimensional image recognition and object reconstruction from CT image sequences for preoperative planning in orthopaedic surgery

The bony structures of the pelvis are automatically recognized from computed tomograms (CT) and are reconstructed to give coherent surface models. The purpose of this procedure is to improve preoperative planning and surgical treatment of pelvic fractures and congenital deformations of the hip joint via interactive computerized simulation of surgical interventions. The process of image recognition is based on the method of moment invariants enabling construction of feature vectors of any magnitude. The elements of a vector are invariant, i.e. independent of the following image transformations: translation scaling and rotation. The vectors, consequently, are largely independent of differences in patient size and positioning during CT acquisition. The computer models, for continued efficient processing of reconstructed parts of the body, are composed of interconnected polygons. These polygonal models are realistically displayed on the computer motor but can be used also for quantitative evaluation of distances, positions, ect. (orig.)