SEGMENTATION OF ARTICULAR FEMORAL KNEE CARTILAGE USING ACTIVE SHAPE MODELS
Open Access
Author:
Durkin, John R
Area of Honors:
Electrical Engineering
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
Dr. David Miller, Thesis Supervisor David Jonathan Miller, Thesis Supervisor Professor William Evan Higgins, Thesis Honors Advisor Kenneth Urish, Thesis Supervisor
Keywords:
osteoarthritis knee cartilage segmentation active shape models active appearance models
Abstract:
Objective: A new method for diagnosing and gauging the progression of osteoarthritis in the knee is using quantitative magnetic resonance imaging. One important step in the process of analyzing the cartilage of the knee for osteoarthritis is the segmentation of the articular femoral knee cartilage. Many conventional segmentation techniques fail because of a lack of signal contrast between the cartilage and soft tissue. The objective of this project was to use active shape models in conjunction with the geodesic active contour algorithm to produce a method for segmenting femoral articular knee cartilage from double echo steady state (DESS) magnetic resonance images.
Methods: Hand segmented femoral articular knee cartilage images from the dataset of the Osteoarthritis Initiative (OAI) were used to create an active shape model. The active shape model was then used to segment femoral articular knee cartilage from DESS images of the knee from the OAI dataset.
Results: Images segmented by the computer are compared to images segmented manually by a physician. The percent pixel-wise difference is calculated.
