A Fully Automated Segmentation of Knee Bones and Cartilage Using Shape Context and Active Shape Models
Behnaz Pirzamanbein · Lund University Publications Student Papers (Lund University) · 2012
In this master's thesis a fully automated method is presented for seg-menting bones and cartilage in magnetic resonance imaging (MRI) of theknee. The knee joint is the most complex joint in the human body andsupports the weight of the whole body. This complexity and acute task ofthe knee joint leads to a disabling disease called Osteoarthritis among theadult population. The disease leads to loss of cartilage and torn cartilagecannot be repaired unless surgical techniques are used. Therefore, one ofthe important parts of nding the disease and planning the knee surgeryis to segment bones and cartilages in MRI.The segmentation method is based on Statistical Shape Model (SSM)and Active Shape Model (ASM) built from a MICCAI 2010 Grand chal-lenge training database. First, all the data are represented by points andfaces. A Shape context algorithm is applied on 60 data sets to obtainconsistent landmarks. The mentioned consistent landmarks and Princi-pal Component Analysis are used to build a Statistical Shape Model. Theresulting model is used to automatically segment femur and tibia bonesand femur and tibia cartilages with Active Shape model. The algorithm istested on the remaining 40 MRI data sets provided by the Grand challenge2010, and compared with six other submitted papers.