A LOCALIZED LEVEL SET METHOD FOR AUTOMATIC SEGMENTATION OF BONE MRI DATA WITHOUT USE OF PRIOR KNOWLEDGE

Fast and accurate segmentation of bone structure from MRI data is a topic of increasing interest, as its applications continue to broaden from direct diagnostic purposes to 3D finite element model, implant design, and pre and intra operative planning [1]. Since 2010, many automatic segmentation methods have been developed and assessed in the grand challenge competition [2]. Prior knowledge based models, such as statistical shape models and atlas based methods seem to have prevailed over pixel based methods. These methods, however, require data set training and may be less suitable for pathologies that are not incorporated in the data set. The aim of this study is to develop and evaluate an adaptive localized level set based method for segmentation of the femur and tibia head in the knee joint without the requirement of prior knowledge.