Novel 3D MRI technique to measure perineal membrane structural changes with pregnancy and childbirth: Technique development and measurement feasibility

Abstract

Introduction and Hypothesis:

We aimed to develop and validate a new MRI-based perineal membrane reconstruction and morphological measurement technique and test its feasibility on nulliparous and parous women in order to determine the effects of pregnancy and childbirth on the perineal membrane.

Methods:

The perineal membrane was traced on high-resolution MRI using 3D Slicer^® and analyses performed using Rhinoceros 6.0 SR23^®. Validation was done by comparing MRI-based perineal membrane reconstruction to dissection measurements in a cadaver. Feasibility of reconstruction was assessed in the following three groups: nulliparous (NP), primiparous women who underwent cesarean delivery (CD), and primiparous women with vaginal delivery (VD). The following parameters were measured: (1) swinging door angle, (2) bony and (3) soft tissue attachment lengths, (4) separation at perineal body level, (5) surface area, and (6) hiatal area. ANOVA and post-hoc comparisons were performed, and the effect sizes (d) were reported.

Results:

Model reconstruction was similar to cadaver dissection findings. Morphological measurements were feasible in all women (NP, n=10; CS, n=6; VD, n=19). Swinging door angle was 13° greater in CD (p-value =.03; d=1.15) and 16° greater in VD (p-value<.001; d=1.41) when compared to NP. VD showed 13% larger separation at the perineal body than NP (p-value=.097, d=0.84) and 23% larger hiatal area than CD (p-value=.14, d=0.94).

Conclusion:

This novel and anatomically validated MRI-based perineal membrane reconstruction technique is feasible. Preliminary findings show that pregnancy and childbirth both influence perineal membrane morphology with VD associated with largest swinging door angle and perineal body separation.

Brief summary:

This novel MRI-based perineal membrane reconstruction technique allows childbirth-related structural changes to be studied, showing increased membrane separation at the perineal body and downward angulation.

Introduction

The perineal membrane (formerly known as the urogenital diaphragm¹) is a structurally important Level III support component involved in hiatal closure.² It is a three-dimensional mass of connective tissue deep in the perineal space that attaches the pelvic organs laterally to the bony outlet and is connected to the levator ani muscles and birth canal.^3–6 To date, our knowledge regarding the effects of pregnancy and childbirth on the perineal membrane structure is extremely limited. Moreover, almost no information exists on structural alterations from normal seen in living women who have given birth or experienced pelvic floor disorders.^7,8

Childbirth-related enlargement of the urogenital and levator hiatuses is well-recognized as the inciting event after which, typically decades later, pelvic organ prolapse develops.⁹ And, although levator ani defects are associated with an enlarged hiatus, levator defect scores explain less than 20% of the variation seen in hiatus size.¹⁰ What explains the other 80% remains unknown.

Theoretically, damage to the perineal membrane may result in altered relationships between pelvic structures and failed hiatal closure.^2,11 Based on our prior anatomical studies^3,7 we propose four potential childbirth-related morphological changes to the perineal membrane: 1. Caudal “swinging door”, 2. Dorsal spreading, 3. Midline separation, and 4. Stretching. To assess these changes, we developed a novel Magnetic Resonance Imaging (MRI) measurement strategy of the perineal membrane that allows us to test hypothesis regarding these potential structural changes.

The aim of this study was to develop, validate and determine the feasibility of our measurement technique for quantifying perineal membrane morphological parameters. We will report our preliminary findings on perineal membrane morphology differences associated with pregnancy and childbirth in three groups of women necessary for future sample size considerations: nulliparous, primiparous with cesarean section, and primiparous with vaginal delivery.

Methods

We conducted a secondary analysis of data from two prior institutional review board-approved pelvic floor research studies. The parent studies were on 1) Nulliparous women and the aging effect on the pelvic floor, and 2) Primiparous women and the childbirth effect on the pelvic floor, the Evaluating Maternal Recovery from Labor and Delivery study.

Study population

MRIs from three groups of women were used for perineal membrane reconstruction and morphological measurement: 1) Nulliparous women (NP), 2) Primiparous women with elective pre-labor cesarean delivery (CD), and 3) Primiparous women with prior vaginal delivery (VD). Women in the VD group all had at least one of the following risk factors for levator ani injury per inclusion criteria in the parent study: length of second stage >150 or < 30 minutes, forceps, maternal age > 35 years, and newborn birth weight > 4000 grams. Women with obstetrical anal sphincter injury were excluded from the VD group due to concerns that this could alter visibility of the perineal membrane (n=14). Women in which the coronal MRI scan did not cover the perineal membrane to its full extent were also excluded (n=27). Women in the parous groups underwent MRI 6 to 8 months postpartum per protocol in the original studies.

MRI protocol

MRI protocols used in the parent studies have been previously published.^12,13 In brief, nulliparous women underwent MRI in the supine position and proton-density-weighted fast spin-echo imaging was performed in the axial, sagittal, and coronal planes using a 3-Tesla Ingenia MRI scanner (Philips Medical Systems, Best, The Netherlands).¹³ For the parous groups MRIs were completed on a 3 Tesla, Philips Achieva, Philips Medical System, Eindhoven, Netherlands and included coronal, axial, and sagittal proton density-weighted sequences.¹²

Reconstruction technique

The perineal membrane was traced on coronal MRI scans (slice thickness 2mm, TE 261, TR 1800) for each subject using 3D Slicer^® software (v 4.5.0, www.slicer.org, Brigham and Women’s Hospital, Boston, MA), based on anatomical landmarks previously established⁷ (Figure 1). Laterally, the ischiopubic rami were used as boundaries for fiducial placement. Medially, at the urethral level, fiducials were placed up to the periurethral tissues where the membrane fuses with the compressor urethrae and urethrovaginal sphincter; at the vaginal level, fiducials were placed up to the outer edge of the vaginal wall. All medial borders placed on coronal images were confirmed with corresponding fiducials on the axial plane. Ventrally, the perineal membrane was marked until corresponding fiducials on the sagittal plane reached the infero-posterior margin of the arcuate pubic ligament. Dorsally, fiducials were placed until the last coronal slice where lateral attachments to the ischiopubic rami could be identified, corresponding to the level of the perineal body. Care was taken not to include branches of the pudendal nerve and/or vessels that enter the perineal membrane at this location. The cranial border was defined by the levator ani muscles and the caudal border marked by the vestibular bulbs and crus of the clitoris.

Figure 1.

Reconstruction of perineal membrane. Left column: left oblique view of pelvis showing perineal membrane array of fiducials (yellow dots) transected by coronal MRI scan at 3 different anatomical levels as labeled. Middle column: coronal MRI scans with perineal membrane tracing (yellow dots). Both left and middle columns illustrate in vivo perineal membrane tracing of the same subject. Right column: coronal section of a 33-year-old female cadaver with normal support showing perineal membrane (between yellow arrows) at each level. Abbreviations: (R) rectum; (V) vagina; (B) bladder; (U) urethra; (VB) vestibular bulb; (CC) crus of clitoris; (LA) levator ani muscle; (OIM) obturator internus muscle; (IRF) ischiorectal fat; (ISC) ischium; (IPR) ischiopubic ramus; (PV) pudendal vessels.

The resulting array of fiducials was imported into Rhino^® (V 6.0 SR23, Robert McNeel & Associates), an engineering software program for 3D surface reconstruction. In Rhino^®, Smooth B spline curves were fitted to fiducials at the attachment boundaries of the bony and soft tissue attachments sites. Next, a smooth contour surface was reconstructed through the perineal membrane fiducials constrained by the boundary curves (Figure 2). The resulting reconstructed surface model was used for quantitative analysis.

Figure 2.

Caudal view of perineal membrane (dark blue) with array of yellow fiducials (panel A) and corresponding surface model (panel B). Structures also seen on panel A: levator ani muscle (light red), external anal sphincter (dark red); perineal body (grey); vagina (tan).

Morphological parameters and potential morphological changes

Figure 3 shows four potential childbirth-related morphological changes to the perineal membrane and the parameters used to quantify those changes.

Figure 3.

Top row: Left oblique view of pelvis with perineal membrane surface model (dark blue) illustrating the four potential childbirth-related morphological changes with normal perineal membrane for reference (semi-transparent blue). Bottom row: Corresponding parameters for each potential morphological change.

Caudal Swinging Door:

In this proposed morphological change, the perineal membrane pivots on its lateral bony attachments to swing downward (caudally).

Swinging door angle (parameter 1) - defined as the angle between the “bony reference plane” and the “perineal membrane plane”, where bony reference plane is the plane through the bony attachment points on the ischiopubic rami and perineal membrane plane is the best fit plane through the fiducials. Because there is a left and right side to the perineal membrane, the average of the swinging door angles from both sides is used. Positive swinging door angle values represent a caudal displacement of the membrane.

Dorsal Spreading:

In this proposed morphological change, the perineal membrane stretches dorsally associated with broader soft tissue attachment length but relatively unchanged bony attachment length. Two aspects were measured:

Bony attachment length (parameter 2) - defined as the lengths of the lateral perineal membrane attachments to the ischiopubic rami. It is measured by fitting a smooth B spline curve to the bony attachment sites on the left and right sides and then calculating the average.

Soft tissue attachment length (parameter 3) - defined as the distance from the arcuate pubic ligament to the superior edge of perineal body. This was chosen over the soft tissue contour to improve measurement reproducibility as a small variation in fiducial placement could significantly affect the curved length.

Midline Separation:

In this proposed morphological change, the ends of the two sides of the perineal membrane are further apart at the level of the perineal body.

Separation at the perineal body level (parameter 4) - defined as the distance between the infero-medial edges of the left and right sides of the perineal membrane at the level of the superior edge of the perineal body.

Stretching:

In this proposed morphological change, the perineal membrane undergoes a global stretching resulting in an increase in surface and hiatal areas. Two aspects were measured:

Surface area (parameter 5) - defined as the sum of the left and right perineal membrane surface model areas.

Perineal membrane hiatal area (parameter 6) - defined as the central space between the left and right sides of the perineal membrane. While this central area contains the urogenital hiatus, perineal membrane hiatal area is a distinctly different parameter from urogenital hiatal area.

Cadaver validation

MRI and dissection of a fresh cadaveric pelvis was performed to validate the perineal membrane reconstruction technique. The specimen came from an 87-year-old nulliparous woman with uterus in situ and without any evidence of prolapse. First, MRI was performed with ultrasound gel in the vagina. High-resolution coronal MRI scans of the specimen were acquired using the same 3T MRI scanner as used for study subjects. Next, pelvic dissection was performed by two co-authors (second and third) who are experienced with pelvic floor anatomy and dissection. The perineal membrane was identified, dissected, and bony attachment length was measured. Bony attachment length was chosen as the validation parameter as this measurement is less subject to tissue distortion inherent to specimen preservation and dissection (Figure 4). The perineal membrane was then reconstructed based on coronal MRI scans of the pelvis and the morphological parameters were measured using the method described earlier. Bony attachment length measurements were compared between the dissection and MRI.

Figure 4.

Cadaver validation of perineal membrane reconstruction technique. In Step, the perineal membrane is traced (yellow dots) on coronal MRI scan of the specimen. In Step 2, tracings are transformed into a surface model (blue). In Step 3, specimen was dissected, and bony attachments measured and compared to surface model parameters. Bottom pictures in steps 2 and 3 correspond to zoomed in images of dashed red rectangles. Abbreviations: (U) urethra; (V) vagina; (IPR) ischiopubic ramus.

Inter-rater reliability

Inter-rater reliability was assessed using 15 subjects - five in each group. Two different raters (first and last authors) independently reconstructed the perineal membrane and inter-rater reliability was determined using intra-class coefficients.

Statistics

ANOVA and Bonferroni post hoc comparisons were used to compare the perineal membrane parameters of all three groups. The percentages of mean difference and the effect sizes were reported. Cohen’s d effect sizes were categorized as large (≥0.80), medium (≥0.50), and small (≥0.20).¹⁴

Results

The perineal membrane reconstruction technique was successfully performed using MRIs from 35 women who were available from the original study that met our inclusion criteria. Subjects were divided in three groups of women according to mode of delivery: NP, n=10; CD, n=6; and, VD, n=19. The NP group was an average of five years younger than the parous groups, and all groups had similar BMI. Demographics and obstetric data for the three groups are shown in Table 1.

Table 1.

Group demographics and obstetric data

Demographics and obstetric data	Nullipara (n=10)	Cesarean Delivery (n=6)	Vaginal Delivery (n=19)	P-value
Age (years)	24.6 ± 3.1	29.2 ± 5.5	29.9 ± 4.9	0.02
BMI (kg/m2)	26.7 ± 4.1	23.9 ± 4.2	25.1 ± 4.3	0.47
White/Caucasian, n (%)	7 (70)	5 (83)	17 (89)	0.79
Newborn weight (grams)	-	3384 ± 462	3527 ± 447	0.52
Newborn head circumference (cm)	-	35.4 ± 2.4	34.9 ± 1.5	0.54
Vaccum delivery, n (%)	-	-	2 (11)	-
Forceps delivery, n (%)	-	-	0 (0)	-
Perineal laceration (1st or 2nd degrees)*, n (%)	-	-	18 (95)	-
Episiotomy, n (%)	-	-	0 (0)	-
Length of 2nd stage of labor (min)	-	-	190 ± 136	-

Data presented as mean ± SD except where noted

^*3rd and 4th not included as anal sphincter laceration was an exclusion criteria

Technique validation with one cadaver dissection showed only a 7% difference in bony attachment length between measures obtained with MRI (4.2 cm) and during dissection (4.5 cm). Inter-rater reliability was assessed with total 15 subjects with five subjects for each group and intra-class coefficients were: .958, .905, .800, .770, .920, and .803 for swinging door angle, bony attachment, soft tissue attachment, separation at the perineal body level, surface area, and perineal membrane hiatal area, respectively.

Swinging door angle (1)

Swinging door angle was statistically different between groups (Figure 5 with numerical values provided in the supplementary table). When compared to the NP group, swinging door angle was on average 13° larger in the CD group (p-value = 0.028, effect size =1.15) and 16 ° larger in the VD group (p-value <0.001, effect size =1.41). No difference was found between the CD and VD groups (p-value >.999, effect size= 0.26).

Figure 5.

Perineal membrane morphological parameters comparison among nulliparous, cesarean and vaginal deliveries groups.

Bony attachment lengths (2)

Bony attachment length was longest in the NP group and was, on average, 13% larger than both the CD and VD groups with effect sizes of 0.86 and 0.83, respectively. The two parous groups had similar bony attachment lengths with only a 0.5% difference (d=0.23).

Soft tissue attachment lengths (3)

Soft tissue attachment length was shortest in the CD group. Compared to nulliparous women, soft tissue attachment length was 14% shorter in the CD group (d=0.85) and 7% shorter in the VD group (d=0.47). This parameter was 7% larger in the VD versus CD group but this difference had a very small effect size (d=0.06).

Separation at the perineal body level (4)

Women with VD had the largest separation at the perineal body level. Compared to NP women, this parameter was 13% larger in the VD group and this difference had a large effect size of 0.84. Group differences were less pronounced between the CD and NP groups with the CD group having a 7% larger separation at the perineal body level (d=0.43) The smallest group difference was seen between the parous groups with separation at the perineal body level being 6% larger in the VD versus CD group (d=0.41). Notably, the VD group had two outliers (exceeding the 95^th percentile) in whom this parameter was 37% larger than the group average.

Surface area (5)

Surface area was largest in the NP group. Compared to NP women, surface area was 13% smaller in the CD group (d=0.62) and 9% smaller in the VD group (d=0.47). Among parous women, perineal membrane surface area was 4% larger in the VD versus CD group (d=0.11).

Perineal membrane hiatal area (6)

Hiatal area was 23% larger in the VD group compared to women in the CD group with a large effect size (d= 0.94). The CD group had on average 17% smaller hiatal area than the NP women with a medium effect size (d = 0.68). VD and NP groups had similar hiatal areas with only a 5% difference (d=0.25).

Discussion

In this study, we established the feasibility of a novel MRI-based technique to quantify structural changes in perineal membrane morphology in vivo. Using this technique, we found evidence suggesting pregnancy and childbirth are both associated with changes in perineal membrane structural morphology. These preliminary results identify swinging door angle and separation at the perineal body level as the most prominent childbirth-related perineal membrane structural changes.

Swinging door angle was the parameter that differed the most between NP and parous groups. This raises the question of whether parity affects swinging door angle by altering perineal membrane morphology leading to a caudal displacement of this structure. Our small sample of women with CD also had a larger swinging door angle compared to the NP group suggesting that this structural change could be an acquired adaptation during pregnancy that does not completely reverse postpartum. However, since women who deliver exclusively by cesarean are less likely to present with pelvic floor symptoms and dysfunction – particularly pelvic organ prolapse,^15,16 an increased swing door angle might solely represent a normal physiological change related to pregnancy and childbirth. The clinical relevance of such a change remains to be investigated in a larger, more diverse cohort. In addition, pregnancy-induced changes in the perineal body warrant further study.¹⁷

The effect of vaginal birth on increasing genital hiatus size is well described^18,19 and was reinforced by our findings. Of note, we did not directly measure genital hiatus length using a traditional clinical measurement but the opening within the perineal membrane (parameter 6) was used as an indirect measure of the genital hiatus. We found the perineal membrane hiatal area to be largest in the VD group, and compared to the CD group, this difference had a large effect size. In addition, we detected a larger separation at the perineal body level in the VD group when compared to NP, again with a large effect size. Interestingly, 10% of the VD women presented as outliers with midline separations 37% longer than the rest of subjects. This raises the question of whether these are a subset of women at increased risk for prolapse or other pelvic floor disorders. Longitudinal studies are needed to answer this question and to determine the impact of the perineal membrane changes identified in the current study on genital hiatus enlargement, as it is a major risk factor for prolapse.^9,20,21 If midline separation of the perineal membrane is proven to be associated with impaired hiatal closure, this parameter could be used for the early identification of women at high risk for prolapse and a novel target for intervention.

Anatomically, this is a complex region of the body with many interconnected parts.³ For example, the medial portions of the levator ani muscle insert onto the cranial surface of the perineal membrane and connections to the perineal body. Now that a technique is available to analyze perineal membrane structural changes, future studies can be conducted to assess how these changes in these different structures interact to determine normal and failed hiatal closure. These factors would include, but not be limited to, muscle innervation, tearing and contractility, as well as perineal body integrity. These findings would need to be investigated in relationship to each of the several pelvic floor disorders. Effect sizes reported in this study can be used to determine sample sizes needed to adequately power future studies.

Our technique allows women’s perineal membrane morphology changes to be studied in vivo for the first time. The method presented was validated in cadaver study and acceptable inter-rater reliability which ensures validity and reproducibility. Feasibility was tested both on nulliparous and parous women, allowing applicability for obstetric injury mechanism studies. This study provides an important tool to further extend our knowledge on how the perineal membrane interacts with the levator ani muscles and the perineal body, how it relates to the hiatal closure mechanism and how it is impacted by childbirth. In addition, the knowledge developed on MRI could then be transferred to ultrasound imaging and applied to postpartum evaluation for the development of preventive measures and early interventions.

Several limitations should be considered when interpreting our study findings. This is a pilot study with a small sample size in a convenience sample, and, therefore, results cannot for now be extrapolated to the general population. Due to the small sample size, some of the morphological differences we detected, even though having meaningful effect size, do not yet reach statistical significance. The VD group include women at high risk for levator ani injury during childbirth, which could limit generalizability. In addition, we did not control for levator ani muscle status in this pilot study, which could affect our results.

In summary, we have successfully developed and validated an MRI-based technique to reconstruct and quantify perineal membrane morphology in vivo. Preliminary findings from this feasibility study point to pregnancy and vaginal birth influencing perineal membrane morphology. This new technique equips us with tools to better study changes in the perineal membrane structure consequent to childbirth and pelvic organ prolapse and to better understand the perineal membrane’s contribution to the hiatal closure mechanism.

Supplementary Material

1819607_Sup_Tab

Supplementary table. Perineal membrane morphological parameters measurements.