Background: This study focuses on osteoarthritis (OA), which affects millions of adults\nand occurs in knee cartilage. Diagnosis of OA requires accurate segmentation of cartilage\nstructures. Existing approaches to cartilage segmentation of knee imaging suffer\nfrom either lack of fully automatic algorithm, sub-par segmentation accuracy, or failure\nto consider all three cartilage tissues.\nMethods: We propose a novel segmentation algorithm for knee cartilages with level\nset-based segmentation method and novel template data. We used 20 normal subjects\nfrom osteoarthritis initiative database to construct new template data. We adopt spatial\nfuzzy C-mean clustering for automatic initialization of contours. Force function of our\nalgorithm is modified to improve segmentation performance.\nResults: The proposed algorithm resulted in dice similarity coefficients (DSCs) of\n87.1, 84.8 and 81.7 % for the femoral, patellar, and tibial cartilage, respectively from 10\nsubjects. The DSC results showed improvements of 8.8, 4.3 and 3.5 % for the femoral,\npatellar, and tibial cartilage respectively compared to existing approaches. Our algorithm\ncould be applied to all three cartilage structures unlike existing approaches that\nconsidered only two cartilage tissues.\nConclusions: Our study proposes a novel fully automated segmentation algorithm\nadapted for three types of knee cartilage tissues. We leverage state-of-the-art level set\napproach with newly constructed knee template. The experimental results show that\nthe proposed method improves the performance by an average of 5 % over existing\nmethods.
Loading....