Segmentation of the bladder wall using coupled level set methods

We describe a novel method to segment the bladder wall in magnetic resonance imaging (MRI) to support the detection of disease, such as endometriosis, and for surgical planning. We segment the inner and outer wall boundary using T2- and T1-weighted MRI images, respectively. A new coupling technique for level sets is formulated and tested on 54 T2- and T1-weighted image pairs. A local phase based dimensionless feature asymmetry measurement using the monogenic signal is used. The results are validated against manual segmentations using the Dice similarity coefficient. Our findings show that the coupling significantly improves the segmentation by preventing leakage due to weak image features and MR bias field. This method shows promising potential for other segmentation tasks involving thin, elongated structures.