Preliminary study for characterizing Legg-Calve-Perthes disease based on MRI segmentation

摘要

Legg-Calve-Perthes (LCP) disease is defined as avascular necrosis of the femoral epiphysis [1] : it particularly distorts the size and shape of the proximal femur and results in disability and abnormal growth of the hip. We choose magnetic resonance imaging to evaluate the physis deformations. In this paper, we present different segmentation methods for depicting the structures of the hip joint. Four algorithms have been implemented : (a) dynamic clustering [2], (b) region growing [3], (c) watershed method [4], (d) snakes [5]. Algorithms have been tested on hip images (T2-weighted, coronal plane) to extract the osseous nucleus and to analyse the shape. A segmentation tool for MR images of hip has been developed. It provides biometric parameters to evaluate the osseous nucleus abnormalities. Further tests have to be made on pathologic images in order to compare with clinical classifications of the disease.