Automatic segmentation of the bony structure of the shoulder
Nhat-Tan Nguyen
Denis Laurendeau (Supervisor)
Problem: Many problems in endoprosthesis placement are due to limited pre-operative insight into the patient's state. Most patients have rheumatoid arthrititis, which affects the quality and shape of the bone: often the glenoid cavity at the shoulder blade has receded. Preoperatively, the operation plan must be devised on the basis of 2-D Roentgen images. During surgery, the operation field is very limited, only the articular surface of the scapula is exposed. The patient is positioned on his side, with the affected shoulder pointing upwards, resulting in a downwards slide of the shoulder blade. During surgery, the most commonly used technique is to remove the remaining cartilage, mill the endoprosthesis shape in the trabecular bone, apply a certain amount of bone cement, place the endoprosthesis and insert screws for early fixation. Others approach the operation field from the back side of the patient by splitting the scapula in two pieces. A preferred technique will support the endoprosthesis on the cortical bone, and try to restore a functional position and orientation. Therefore, a custom-sized prosthesis has to be used, and a good view of the shoulder blade is needed.
Motivation: Automatic segmentation of bony structures in MRI angiography datasets is an essential pre-processing step necessary for most visualization and analysis tasks. In order to have a better understanding of the healthy and pathologic arthrosis shoulder, this project will develop a method to segment the structures of the shoulder from MRI datasets.
Approach: Finding suitable techniques for segmenting the scapula and humerus in a bidimensional MRI data-set is a major part of this project. Currently we focus on contours-based segmentation techniques, i.e., first a point of the bone is identified. The bone boundaries are extracted in a second step.
Calendar: September 2003 - September 2005
Last modification: Sep 28 2007 2:44PM by ntnguyen


©2002-. Computer Vision and Systems Laboratory. All rights reserved