Kinematic analysis of the lumbar spine from fluroscopic images

Spine research grew with the development of medical imaging techniques. Using the common x-ray radiography the range of spinal movement could be studied. Later, the development of Digital Videofluoroscopy made possible the acquisition of real-time images of the moving spine. However, in order not to expose the subject to a high dose of x-radiation, the quality of the images is rather poor.Kinematic analysis is based upon anatomical landmarks placed on the vertebral images. In this thesis an improved landmarking procedure is presented. It is a combination of two algorithms. The first one is cross-correlation (a method applied in most of the previously developed landmarking algorithms) and is used to locate the corner of the vertebral body. The second is a new algorithm (called the First Pixel algorithm) which is used to locate the landmark on the contour of the vertebral corner. The procedure was tested both on DVF images of a calibration model and in vivo images. An inter- and intra-observer study was carried out. Using the new algorithm the repeatability is increased by 85% with respect to manual landmarking.A new geometric method to compute the location of the instantaneous centre of rotation (ICR) was developed in order to cope with angular displacement smaller than 3°-5°. Using the new method, the influence of vertebral translation upon the location of the ICR and disc bisector intersections was studied using landmarks obtained from DVF images of a calibration model.Based on the characteristics of the ICR and disc bisectors intersection distribution, two methods to quantify vertebral translation have been developed. Each of the methods has advantages and drawbacks. However, they complement each other and therefore the two methods are used together. The first one establishes the direction of the movement while the second determines the amount of the vertebral translation.