Quantification of 2D shape variation of abnormal proximal femurs over normal proximal femurs in bilateral hip MRI: A computerized analysis for Legg-Calve-Perthes disease

Abstract

This article introduces a novel computerized quantitative shape variation analysis of the abnormal proximal femurs in magnetic resonance imaging (MRI) slices for the Legg-Calve-Perthes Disease (LCPD) patients. The morphological shape variations of the abnormal proximal femurs are analyzed by using the shape information of the normal proximal femurs that are in the same slice as the abnormal proximal femurs. In order to measure the shape variation of the abnormal proximal femurs, the healthy and pathological proximal femurs in bilateral hip MRI slices that are obtained in coronal plane are segmented manually in the first step. Secondly, the image regions of interest (RoI) of the manually annotated proximal femurs are extracted from the entire segmentation by image cropping. Then, the cropped images of the normal and abnormal proximal femurs are aligned automatically with rigid image registration. Finally, the shape variation of the abnormal proximal femur is quantified by measuring the percentage of the area outside the intersection over union of the aligned proximal femurs. In the scope of the proposed study, a total of 13 MRI slices of 13 LCPD patients whose one proximal was normal and the other one was abnormal were evaluated. In quantification of the morphological variations of the abnormal proximal femurs, shape variation ratios of 09.45% (± 01.40), 18.27% (± 06.30), 24.73% (± 12.42), 27.59% (± 10.27) were observed for the four phases of LCPD as Initial (3 patients), Fragmentation (5 patients), Reossification (2 patients), Remodelling (3 patients) respectively. The computerized shape variation analysis discussed in this article may be useful for clinical evaluations of the patients with LCPD. Moreover, proposed approach can be performed in computerized analysis of the other hip disorders that affect proximal femurs.