| Περίληψη: | Dynamic assessment of three-dimensional (3D) joint kinematics is
essential for understanding normal joint function as well as the effects of injury
or disease.
The knowledge of one or two series of bi-dimensional fluoroscopic projections of the joint in motion (mono-planar or bi-planar fluoroscopy), and the 3D model of the joint segments, were claimed to be sufficient to reconstruct the absolute and relative 6 Degrees Of Freedom (DOFs) pose of bones or prostheses in the 3D space.
The software MultiTrack was developed at the Health Sciences and
Technologies - Interdepartmental Center for Industrial Research (HST - ICIR) for
the joint kinematics estimation with 3D Video Fluoroscopy (3DF) [1] using C++
language with ITK [2] segmentation & registration toolkit and VTK [3]
visualization toolkit.
An optimization procedure finds the 6 degrees of freedom pose that optimizes a metric quantifying the matching of the 3D model and its relevant projections. The metric, currently implemented in the software, is based on the contour segmentation of the object to be tracked and on the use of 3D adaptive distance maps (ADM) [4,5]. However, the contour extraction is a time consuming procedure for the user. Different methods were proposed in the literature to reduce the user interaction, each with its proper pros and cons.
In the current thesis a few of the for-mentioned methods are discussed in order to evaluate each of them in terms of accuracy, speed and user dependency. At the final step the algorithm proposed by Mafhouz et al. [6], initially proposed for prostheses, is implemented inside the MultiTrack framework. To be properly characterized, the above method is tested on in vivo datasets and under various sources of error.
|
|---|
