Segmentation of Magnetic Resonance Images from Female Pelvic Cavity

Magnetic resonance imaging is currently one imaging modality for studying pelvic floor dysfunctions. In order to perform biomechanical analysis, the geometrical models of the concerned structures are needed, which implies that these structures should be segmented in the acquired image series. However, the appearances of the organs and muscles of female pelvic cavity can be easily distorted in the images by noise and partial volume effect, which leads to the failure of common segmentation algorithms. In this study, we propose algorithms to handle the segmentations of the pelvic organs and muscles in T2-weighted axial magnetic resonance images. The proposed algorithms are based on the imaging features of different structures, and use various image clues and prior knowledge for the segmentation. Implementation details and further issues are introduced and discussed. Additionally, numerical examples are included to demonstrate the effectiveness of the proposed algorithms.