Masters Thesis, Boise State University, December 2009
Fluoroscopic analysis of knee joint kinematics involves accurately determining the position and orientation of bones in the knee joint. This data can be derived using the static 3-D CT scan images and 2-D video fluoroscopy images together. This involves generating hypothetical digitally reconstructed radiographs (DRR) from the CT scan image, with known position and orientation and comparing them to the original fluoroscopic frame. This represents a search problem in which, among all the DRRs possible from a CT image, the image that matches the target fluoroscopy frame of the knee joint has to be searched.
Each image in the search space differs from another by the set of position and orientation values. Position is defined by the x, y and z co-ordinates in the Cartesian coordinate system. Orientation is defined by the values of azimuth, elevation and rotation. Therefore this constitutes a six dimensional search problem. Using a brute force method to search for the target image which is associated with six independent values, each of which has a large range of possible values, can take a tremendous amount of time. The fact that the set of six values cannot be ordered or categorized in any way adds to the complexity of the search. Further, previous research conducted by Charles Scott et al.  suggests that even good sequential search algorithms such as the sequential Monte Carlo method can be very time-consuming. Therefore, a search solution which is suitable for this kind of 6-D search and one that provides a significant speedup has to be used. This thesis involves using Swarm Intelligence (SI) techniques in a parallel computing environment to achieve faster results. Parallel programs are developed using two SI techniques, Bees Algorithm and Particle Swarm Optimization technique.