Variations in Levator Ani volume and geometry in normal living subjects, versus prolapse, and genuine stress incontinence: The application of MR based 3D reconstruction in evaluating pelvic floor dysfunction

Objective: We report on the comparative 3-dimensional (3D) living female pelvic floor geometry in five women, comparing the volume, morphology, and integrity of the levator sling, and pelvic anatomic relationships among study subjects. Materials & Methods: Five women of varying ages, parity, continence, and prolapse status were studied. Two dimensional (2D) imaging of the pelvic floor organs was performed on each subject in the supine position. Manual segmentation techniques and solid modeling software was used to build 3D models of each patient's pelvic floor structures, which could then be viewed and measured on the computer screen. We measured levator muscle volume, posterior urethro-vesical angle, distance from the urethra to pubococcygeal line, and the levator plate angle. The integrity of pubo-coccygeal attachments was also recorded. Results: Levator muscle volume ranged from 68 ml in the nulliparous female, to 26ml in the grand multipara with severe prolapse and mild GSI. The second lowest volume (30ml) was in the multipara with GSI. Volumes in the parous subjects without SUI or pelvic organ prolapse were 36 and 39ml. Pubococcygeal attachments were found to be torn in the 2 symptomatic subjects, and were intact in all 3 asymptomatic subjects. Conclusion: MR based 3D modeling is feasible and can be used in a research setting to evaluate complex anatomic relationships which may accompany pelvic floor dysfunction. The technique can also be used to evaluate levator muscle morphology and volume, as well as pelvic floor support integrity and its possible role in GSI and prolapse. We are currently conducting a larger study to validate our technique, and to better define the relationship between pelvic floor geometry and pelvic floor dysfunction. Introduction Although many studies have been performed to unravel the mechanisms underlying the genesis of prolapse and genuine stress incontinence due to hypermobility (GSI), the role of the levator ani muscles remains poorly understood. Nonetheless, the levators remain the focus of many non-surgical therapies for the treatment of this condition, but with varying degrees of success . Female stress urinary incontinence and pelvic floor prolapse is thought to be due to levator ani muscle, fascial and and/or nerve damage incurred during childbirth trauma . When conservative therapy for genuine stress incontinence (GSI) fails, surgical intervention is aimed at re-supporting the bladder neck . Commonly employed surgical procedures include the Burch colposuspension , bladder neck sling , needle urethropexy , and the Kelly plication . These procedures have reported complication rates of 6-32% , and failure rates of 3-40% 7, . Various surgical procedures have been advocated for repair of pelvic organ prolapse depending on the site of the defect 9 . Despite surgery, prolapse can recur in up to 34% of cases . It is possible that the recurrences may be due to failure to identify the causative lesion in the pelvic floor prior to surgery. The causative lesion could be a site-specific fascial defect or even anatomic disruption of the levator ani muscles. Conventional 2D magnetic resonance imaging (MRI) has been used to assess the anatomy of the female pelvic floor. Strohbehn, et al., reported excellent correlation between cadaveric dissection and MRI findings . In living women, MRI has also been employed to correlate imaging findings with the presence of clinical pelvic floor dysfunction. Huddleston, et al., reported three alterations in vaginal shape that were associated with pelvic floor prolapse . Kirschner-Hermanns, et al., reported increased T1 signal intensity as evidence of muscle atrophy in 66% of their subjects with stress incontinence . Healy used MRI to show a statistically significant increase in cystoceles and recto-uterine prolapse in constipated subjects when compared to normals and those with fecal incontinence . When we investigated the value of MRI of the pelvis in the sitting position, we found a trend towards levator muscle laxity and thinning in women with stress urinary incontinence . However, it is possible that our results were a reflection of the wide normal range of levator muscle volume, and this bears further study. The disadvantage of 2D MRI stems from its inability to easily disclose the 3 dimensional relationships which may be at the root of the defects which lead to clinical pelvic floor pathology. To better understand the specific anatomic defects that may be at issue here, we evaluated the morphology and volume of pelvic floor support muscles, and the bladder neck, using three-dimensional MR based models in 5 living women, with and without clinical pelvic floor pathology. The immediate goal was to demonstrate the feasibility of the technique of 3D pelvic floor modeling in various clinical settings. Materials and Methods We evaluated a small group of subjects with a wide range of age, parity, prolapse, and continence status. One subject was chosen from each category. Inclusion criteria consisted of willingness and ability to tolerate an MRI scan, and ability to give informed consent. Exclusion criteria were prior pelvic or vaginal surgery (cesarean delivery, tubal ligations and laparoscopy excepted), and ferromagnetic implants. Accordingly, five women were selected to participate: three women without urinary incontinence; 1) a 35 y/o nullipara, 2) a 29 year old primipara, 3) a 29 year old multipara, and two with incontinence; 4) a 44 year old multipara with GSI, and 5) a 75 year old grand multipara with grade 4 vaginal vault prolapse and occult GSI. All subjects were asked if they had urine leakage with coughing or laughing. In those who answered affirmatively, diagnoses of GSI were made with the aid of multichannel urodynamic evaluation. Conventional 2D MRI studies were performed: T2-weighted source images were obtained in the axial and sagittal planes using a 1.5T magnet (General Electric Medical Systems, Milwaukee, WI) and a torso phased array coil wrapped around the pelvis. The following imaging parameters were employed: TR=4200msec, TE(eff) =108msec, 128 phase encodes, 24cm field of view, 3mm slice thickness, no gap, 2 acquisitions. The entire sequence was repeated adjusting the slice locations to obtain contiguous images 1.5mm in thickness. Total scanning time was approximately 19 minutes. After the MRI acquisitions were completed, data was electronically transferred to a Sun UltraSparc-30 graphics computer workstation (Sun Microsystems, Mountain View, Ca). The data was first segmented into anatomically significant components, including bladder, urethra, uterus, vagina, rectum, muscles, and bones and then labeled using a combination of semi-automated and manual editing. The manual segmentation time was approximately 10 hours per subject. From these images, 3D renderings of the pelvic viscera as well as supporting muscle, fascia and bones were reconstructed using the marching cubes algorithm and a surface rendering method . Three-dimensional surface models were generated using a pipeline consisting of dividing cubes, triangle reduction, and triangle smoothing 17 . The computer time for 3D model generation was approximately 1.5 hrs per subject. Volumes were computed by multiplying the number of volume elements (voxels) enclosed by the segmented 3D volume, by the voxel volume, which is the product of the picture element (pixel) area (0.8789 square mm in our series), and the effective slice thickness (1.5mm). To compute linear measurements, the user first defines a path of interest. The 3D Cartesian components (i.e.., x, y, z) of the path are determined, and the resultant vector ( R ) is the square root of the sum of the squares of each of the Cartesian components: R = ((x+y + z ( Angular measurements are obtained by defining the angle on the computer screen, and its value is computed by trigonometric analysis. The integrity of the pubo-coccygeal attachments is determined by measuring the Pubo-Coccygeal Gap (PCG), i.e., the distance (in cm) from the most anterior aspect of each arm of the muscle to the nearest inferior aspect of the symphysis. The final results were viewed on a workstation with graphics acceleration, and specialized measurement software was used to compute the linear and volume measurements of the on screen renderings. Representative 2D axial slices are given in Figure 1a and b. Examples of the 3D renderings are shown in Figures 2 and 3, including definitions of the measurement parameters Results The subjects tolerated the MRI scans well. Good quality source images were obtained, and we were able to identify most of the structures of interest easily on the MR source images. The one exception was the gap in the pubic attachment of the pubococcygeus, which was difficult to identify completely in its most anterior-superior aspect from the axial slices. In this case, we traced the muscle as far anterior as it could be reliably identified. Segmentation of the structures, and conversion to 3D was routine. Computation of volume, linear, and angular measures were accomplished interactively, and recorded on a standardized form. Data storage required approximately 20Megabytes per subject, and was found to be adequate. Quantitative results are given in Table I for the 3D models and Table II for the source images. Computed bladder volumes were 23 and 30.5 cc in two of the subjects without GSI. In the nulliparous non-GSI subject, bladder volume was 148cc, but this elevated volume was probably due to her failure to void as instructed prior to the MRI study. In the GSI only patient, bladder volume was higher at 57cc, and it was 136 cc in the subject with prolapse and occult GSI, the latter likely due to urinary retention caused by the buttressing action of a 10cm uterine fibroid which kinked this subject's bladder neck. The length of the PCL varied from 87mm in the nullipara, to 107mm in the asymptomatic para3, without correlation to symptomatology. The distance from bladder neck to PCL was 13mm and 15.9mm in the nullipara and primipara, but increased to 22 23 mm in the multiparous subjects. The posterior urethro-vesical angle varied from 125 degrees in the asymptomatic multipara, to 173 degrees in the grand multipara with prolapse, without apparent correlation to symptomatology. The levator plate angle (LPA) was the most difficult parameter to estimate, owing to the curved shape of the median raphe, and we found that this parameter varied between 12 and 19 degrees in all subjects except the 29 year old primipara, where it was 38 degrees. The volume of the levator ani muscles ranged from 68 ml in the nulliparous female, to 26ml in the grand multipara with severe prolapse and occult GSI. The second lowest volume (30ml) was in the multipara with GSI. Volumes in the parous subjects without stress urinary incontinence were intermediate at 36 and 39 ml. Our marker for pubic ligament attachment defects is the pubo-coccygeal gap (PCG). The measurements fell into 2 groups; less than 2cm in the 3 asymptomatic subjects, and >3cm in the 2 subjects with symptoms. Larger gaps (>3cm) may suggest torn ligaments and shorter gaps (<2cm) suggest an intact ligament. The prolapse subject showed a gap of 5.4cm on the right (suggestive of a tear) and 2.3cm on the left (borderline), and the GSI subject showed a gap of 3.1cm on the left, (suggestive of a tear), and a contralateral gap of 1.7cm (suggestive of an intact ligament). The asymptomatic patients all show gaps less than 2cm. Finally, the maximal width of the levator hiatus was examined. With the exception of the 62mm finding in the grand multipara with prolapse, this parameter varied minimally in the 32 39mm range.