C. Barillot, D.R. Haynor, and P. Hellier (Eds.): MICCAI 2004, LNCS 3216, pp. 258–265, 2004. © Springer-Verlag Berlin Heidelberg 2004 Statistical Shape Modelling of the Levator Ani with Thickness Variation Su-Lin Lee 1,2 , Paramate Horkaew 1 , Ara Darzi 2 , and Guang-Zhong Yang 1,2 1 Royal Society/Wolfson Foundation Medical Image Computing Laboratory 2 Department of Surgical Oncology and Technology Imperial College London, London, UK {su-lin.lee, paramate.horkaew, a.darzi, g.z.yang}@imperial.ac.uk Abstract. The levator ani is vulnerable to injury during childbirth and effective surgical intervention requires full knowledge of the morphology and mechanical properties of the muscle structure. The shape of the levator ani and regional thickening during different levels of physiological loading can provide an indication of pelvic floor dysfunction. This paper describes a coupled approach for shape and thickness statistical modelling based on harmonic embedding for volumetric analysis of the levator ani. With harmonic embedding, the dynamic information extracted by the statistical modelling captures shape and thickness variation of the levator ani across different subjects and during varying levels of stress. With this study, we demonstrate that the derived model is compact and physiologically meaningful, demonstrating the practical value of the technique. 1 Introduction The pelvic floor is a complex structure composed of a diaphragm of striated muscle covered by fascia. The levator ani is the part of the pelvic floor for visceral support and for drawing the anal canal upwards and forwards. It is highly susceptible to injuries during natural childbirth, which can lead to problems with urination and defecation [1]. Defective pelvic organ support is one of the most common problems in women. Among them, over 20% require a second operation. The management of pelvic floor disorders in women is traditionally divided between gynaecologists, urologists and colorectal surgeons. The modern surgical approach to the problem is to perform all necessary repairs during one single operation to avoid the usual territorial subdivision and patient re-operation. This requires an accurate preoperative identification of all prolapsed organs and their mechanical properties. Magnetic Resonance Imaging (MRI) in both closed and open access scanners has been used to assess any injuries of the levator. Open access (interventional) scanners are becoming popular as the subject sits in an anatomically natural position, where the pelvic floor muscles can optimally exert its full function [2]. Dynamic information can also be acquired by using fast acquisition protocols with which each image can be acquired in less than 2 seconds. Full 3D acquisition of the entire levator ani volume during straining has so far proven to be difficult due to the prolonged imaging time required to recover the fine structural details of the levators. Diagnosis has usually