Development of a Combined 2D-3D Image Registration Method for Dynamic Measurement of Knee Joint Tracking

Please enable javascript in your browser.

Sobhanieh, Niloufar | 2014

819 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 46054 (08)
University: Sharif University of Technology
Department: Mechanical Engineering
Advisor(s): Farahmand, Farzam
Abstract:
2D-3D image registration has been adopted in many clinical applications such as kinematic analysis of bones in knee and ankle joints. Because it is not possible to take dynamic images in 3D and 2D dynamic images such as fluoroscopic images lack spatial information. For comparing two images it’s necessary to make their dimensions the same. So in this research we prepared a code in MATLAB to generate digitally Reconstructed Radiograph (DRR) from the CT image in different positions and then be compared with 2D fluoroscopic image. Registering two images could be done either with comparing their intensities or their mutual features. We combined these two methods and developed a cost function with includes both of these. For optimizing the comparing function, Powell’s method is used. We applied this combined method on some data and evaluated the results with foreknown position of DRRs. The results show that this method has the error of less than 1 degree in rotation and 1 millimeter in transposition. Period of the registration depends on resolution of images
Keywords:
Optimization ; Multiresolution ; Intensity Bias ; Hybrid Methods ; Knee Joint ; Digitally Reconstructed Radiograph (DRR)

No TOC