Technique Development and Measurement of Cross-Sectional Area of the Pubovisceral Muscle on MRI Scans of Living Women

Abstract

Introduction and Hypothesis:

Measurements of the anatomic cross-sectional area (CSA) of the pubovisceral muscle (PVM) in women are confounded by the difficulty of separating the muscle from the adjacent puborectal (PRM) and iliococcygeal (ICM) when visualized in a plane orthogonal to the fiber direction. We tested the hypothesis that it might be possible to measure the PVM CSA within a defined region of interest based on magnetic resonance images (MRI).

Methods:

MRI scans of eleven women with unilateral PVM tears and seven primiparous women with intact muscles following elective C-section were used to identify the PVM injury zone defined by the mean location of its boundaries with the adjacent intact puborectal and ICM from existing anatomic reference points using 3D Slicer and ImageJ software. Then, from the 15 or more 2 mm transverse slices available, the slice with the maximum anatomic CSA of the left and right PVM was found in 24 primiparous women with bilaterally intact muscles who had delivered via C-section.

Results:

The mean (± SD) of the maximum left or right PVM cross-section areas for the 24 subjects, measured by two different raters, was 1.25 ± 0.29 cm² (range 0.75- 1.86). The 5th, 50th and 95th percentile values were 0.77, 1.23 and 1.80 cm² respectively. The inter- and intrarater measurement repeatability intra-class correlation coefficients exceeded 0.89 and 0.90, respectively.

Conclusions:

It is possible to use MRI to identify the volume of interest with the maximum anatomic cross-section of the PVM belly, while minimizing the inadvertent inclusion of adjacent PRM or ICM in that measurement.

Introduction

The levator ani muscles form the pelvic ‘floor’ that supports the abdominal and pelvic organs within the pelvis. In the female this floor contains a hiatus through which the pelvic organs pass and, unfortunately, through which the pelvic organs can prolapse (for review, see [1]). The pubovisceral portion of the levator ani (also known as the puboccocygeal muscle) originates bilaterally from the pubic rami on either side of the pubic symphysis and inserts into the lateral walls of the pelvic organs and perineal body [2]. It maintains a constant muscle tone to maintain hiatal closure, providing additional tone to ensure this during increases in intraabdominal pressure [3].

MRI examination of the levator in women with and without prolapse has shown that 55% of women with pelvic organ prolapse display evidence of a major obstetrical tear (defined as >50% missing muscle) in the pubovisceral muscle (PVM) compared with only 16% in women with normal support [4]. Using ultrasound, Dietz and Simpson found similar results and reported that 60% of women with an uterine prolapse have some degree of tear, but only 14% of women without prolapse [5]. It is the impaired action of this muscle that renders it incapable of counteracting the forces imposed on the hiatus by abdominal pressure, an imbalance that can eventually result in prolapse of the pelvic organs [6]. However, a woman with a naturally bulky PVM might lose 25% of her muscle to a type I tear [7] but still have more muscle than a woman with an intact but inherently small muscle. So, it seems desirable to be able to accurately quantify the amount of PVM muscle that is present in a given individual. Functionally, whether there is a tear or not, it is the amount of muscle, specifically the maximum anatomic cross-sectional area of the muscle perpendicular to the fiber direction, on that side of the body that determines how much force that muscle can generate.

Both ultrasound and MR images have been used in the past to measure PVM ‘thickness’ [8,9]. Muscle physiologists always estimate the maximum force-generation capability of a parallel fibered striated muscle under isometric conditions by measuring its anatomic cross-sectional area (CSA) taken perpendicular to the fibers [10] For example, a hypothetical rectangular cross-sectioned muscle that is 1 cm thick by 2 cm wide (in other words having a CSA of 2 cm²) will develop twice as much isometric force as a 1 cm thick muscle that is 1 cm wide (CSA: 1 cm²). So, one can see that thickness alone is not sufficient to know a muscle’s force-generating capacity. That is why we will measure anatomic cross-sectional area not thickness. In addition, the CSA needs to be measured in the muscle “belly”, at the point where the CSA of the muscle is largest (Figure 1). Although PVM CSA has been estimated by first making a 3D model of the muscle and then sectioning it [11], this laborious process introduces potential errors at many steps in model construction. It would therefore be preferable to make CSA measurements from the original MR images, not the model. In addition, the problem has not been addressed that in several regions the PVM is overlapped by the puborectal or iliococcygeal muscle thereby complicating measurement of PVM CSA.

Figure 1.

i) Lateral view of the female pelvis showing the pubic symphysis (PS), bladder, uterus, vagina, rectum and anal sphincter. The pubovisceral muscle (PVM), iliococcygeal muscle (ICM) and puborectal muscle (PRM) are shown with their distinct line-of-action directions. Point A represents the arcuate pubic ligament and point B the anterior wall of the perineal body.

ii) The black box represents the extent of the natural muscle boundaries, including the ventral (19 mm) and dorsal (11 mm) boundaries of the PVM, avoiding overlap with the PVM and ICM. ATLA denotes arcus tendineous levator ani. The red arrow indicates the point at which the puborectal overlaps the pubovisceral muscle. The dotted red lines represent the ICM and PRM boundaries.

iii) The PVM volume of interest is shaded in red and ‘a’, ‘b’ and ‘c’ represent selected transverse planes of this volume. Cross section ‘b’ is the maximum cross-sectional area.

The goal of this paper, therefore, was to develop a technique to measure the maximum CSA of the muscle subdivision in the PVM injury zone in women with or without muscle tears directly from MR images and report on the range variation found in its CSA in a cohort of 24 primiparous women, along with inter- and intrarater reliability . We tested the hypothesis that it is possible to identify the extent of the injury zone in 3-D from existing landmarks in a convenience sample of women with unilateral tears, thereby yielding the dimensions of the region of interest within which to seek the maximum anatomic CSA of the PVM directly from the MR scan. We also tested the hypothesis that the coefficient of variation in the measured PVM CSA is similar, at 25%, to that found in other small, parallel-fibered, striated muscles in the body.

Methods and Materials

This is a secondary analysis of a convenience sample of all women with the unique characteristics of a unilateral PVM tear or having had a C-section present in our database. They represent women with unique and specific characteristics, so the sample size was determined by what was available. We were not comparing measurements, so a power calculation was not part of our study design. Eleven parous women with a complete unilateral levator ani tear were identified in whom the intact or slightly injured levator muscle present on one side could directly be compared with a damaged levator muscle on the other side (i.e., women with levator score 3-1 and 3-0 were included). Assuming that it is mostly the PVM rather than the puborectal muscle that is injured during birth [12] (see also Discussion) the demarcation between these two overlapping muscles was evaluated by comparing the intact side with the contralateral side with missing PVM. To do this we compared the side that contained the intact PVM, puborectal and iliococcygeal muscles with the side on which the PVM was missing. The portion of the muscle that was seen on the intact, but not the injured, side allowed us to define each boundary between this muscle and the two adjacent muscles, namely the iliococcygeal muscle (ICM) and the puborectal muscle (PRM). These scans were available from a case-control study (“OPAL II”) of women with pelvic organ prolapse with protrusion below the hymen and no prior surgery for prolapse. Women who served as controls were asymptomatic research volunteers with normal support (i.e., no prolapse to the level of the hymen) upon examination; they were recruited to be of similar age, parity and hysterectomy status. Because, at least theoretically, intact muscle position might be altered by the injury on the contralateral size, we examined women who had not given birth vaginally. Scans were available from the EMRLD study (Evaluating Maternal Recovery from Labor and Delivery) from 24 primiparous women 8 months post-partum who delivered by cesarean section. They had normal muscles not altered by vaginal birth and neither had a PVM tear nor prolapse on POP-Q examination: seven women who had an elective C-section were used for method development and all 24 women who had either an elective or with active 2^nd stage of labor C-section were used to calculate PVM CSA (Table 1). A secondary analysis of a convenience sample was performed instead of the recruitment and scanning of a new set of women due to the cost associated with the additional MRI’s. Both studies were approved by the University of Michigan Institutional Review Board. These scans were used for the purposes of obtaining the CSA of normal muscles.

Table 1 –

Overview of subject participation by protocol phase, method development: boundary identification and measurements and inter and intra-rater reliability, indicated by ‘y’ (yes). Each rater performed both the 1^st and 2^nd measurement. Only data from the first measurement are reported.

Group	n	Method development: boundary identification	PVM CSA measured?	Measurements and inter and intra-rater reliability
				1st Measurement	2nd Measurement
Unilateral Tear	11	y
C-section
Elective C-section	7	y	y	y	y
C-section with 2nd Stage	17		y	y	y

MRI images were obtained with a 3-T scanner (Achieva^™, Philips Healthcare) with an eight-channel cardiac coil positioned over the pelvis. They included coronal, axial, and sagittal proton density-weighted (TR/TE, 2,100–2,500/30 ms) sequences (field of view, 20 cm; matrix, 256 × 25; number of signal averages, 2; slice thickness, 4 mm; gap, 1 mm). Additional 2-mm sections with 0.2-mm gaps in the axial and coronal planes were obtained with proton density-weighted sequences (TR/TE, 2,100–2,500/30 ms) [13]. The 2 mm scans were used primarily for the cross-sectional measurements and the others for navigating general pelvic architecture. The MR images were imported into 3D Slicer (v. 4.5.0), so slices containing the volume of interest could be identified in a plane perpendicular to their muscle direction. The captured images were then exported into Image J (v.1.50i) for measurement because measurements are not possible in 3D Slicer in reformatted planes.

Defining the PVM boundaries

To identify a plane perpendicular to the PVM, its line-of-action was first established as a plane between origin and insertion, namely the inner surface of the pubic bone and its insertion into the perineal structures. The boundary between the PVM and puborectal muscles (“PRM boundary”) and the PVM boundary with the iliococcygeal muscle (“ICM boundary”) were then identified (Figure 1; Figure 2 i and ii). The following protocol was used to examine the PVM in women with a unilateral tear in which one side had PVM and the other was missing (Figure 2 iii); note that both the PRM and ICM could be seen (Figure 2 iv-v):

1. Identify PVM line-of-action in a parasagittal section using the following factors: 1) the direction of the visible striations within the muscle (Figure 3 i) and knowledge that 2) the muscle passes from the inner surface of the pubis to the vagina, perineal body and anal sphincter complex, and 3) that the lower boundary of the muscle lies against the perineal membrane.

2. Rotate the coronal plane to lie perpendicular to PVM line-of-action.

3. Identify PRM boundary by moving the plane back and forth and comparing the side with missing PVM muscle to the side with intact muscle. The boundary was considered to occur where PRM muscle mass is seen on the side where the PVM was absent (Figure 2 iv). The perineal body was selected as a reference structure for boundary location measurements because it was consistently present and visible in all scans whether or not a defect was present.

4. Identify the anterior aspect of the perineal body in a mid-sagittal plane. Then, find its location in the coronal plane and measure its distance from the identified PRM boundary.

5. ICM boundary: Using the same rationale, identify the ICM boundary by moving the plane back and forth comparing the side with missing muscle with the side with intact muscle (Figure 2 v). The arcuate pubic ligament was chosen as the reference structure from which measurements of boundary location are made.

6. Using the first scan where the arcuate pubic ligament is seen when moving posterior-anterior. Place a ruler perpendicularly from the arcuate pubic ligament to the identified boundary and measure the distance.

Figure 2:

MRI of healthy women (panel i, ii, vi, vii) and of a woman with unilateral tear (panel iii-v). Panels iii-vii are in the coronal tipped plane.

i) 3D representation, pubic symphysis (PS), levator ani (LA) and coronal plane rotated to the angle normal to the fiber direction (tipped plane).

ii) 3D representation of LA and coronal tipped plane.

iii) Healthy pubovisceral muscle (PVM) on the right side, with absence of muscle on the left side of the image (both features indicated by arrow heads). PBo – Pubic Bones, Ut – Uterus, U – Urethra, OI – Obturator Internus.

iv) When moving back and forth on the coronal tipped plane, there is a point where the puborectal muscle (PRM) is visible and intact on both sides. This point is located 11 mm ventral to the anterior aspect of the perineal body.

v) The PVM is absent on the left-side, but visible on the right side (filled arrow head) is visible on both sides (arrow heads). The ventral boundary of the PVM is located 19 mm ventral to the arcuate pubic ligament.

vi) Puborectal muscle in a healthy individual (arrow heads), muscle morphology is different than the one seen in vii. This subtle difference allows the separation between both portions of the levator ani in healthy individuals.

vii) PVM transition to iliococcygeal muscle (ICM) in healthy individual (white arrow heads), seen as a small indentation cranial to the PVM belly.

Figure 3:

MRI of healthy woman showing the different steps used to calculate the PVM (pubovisceral muscle) CSA (cross-sectional area).

i) Parasagittal plane, showing the PVM direction (green line). PS – Pubic Symphysis, B-Bladder, Ut – Uterus.

ii) 3D view of the sagittal plane and the tipped coronal plane, rotated to an angle normal to the PVM direction.

iii) Coronal plane, with the tracing of the ventral limit of the PVM, located 19 mm ventral to the arcuate pubic ligament (APL).

iv) Midsagittal plane, showing the location of the anterior aspect of the perineal body (PB).

v) Tipped coronal plane moved to be 11 mm ventral to PB.

vi) Tracing of the PVM CSA.

Because the absence of muscle unilaterally might lead to a distorted anatomy we then applied this technique to women that had undergone elective C-section whose muscles were normal on MRI. Using the insights gained from the unilateral set, it was possible to identify anatomical features identifying the boundaries in the intact muscles based on visible features. These were: 1) for the PRM boundary it was noted that there was a different fiber organization in the muscle belly (Figure 2 vi) and 2) for the ICM boundary a small indentation was seen in the muscle belly that “separates” the ICM from the PVM (Figure 2 vii).

Next, we developed a measurement system that would first identify the PVM injury zone that precluded inadvertently overlapping from the ICM and PRM. The distances of both PRM and ICM boundaries to the anatomical reference points were measured for all subjects used in the method development phase (Table 1). The mean distance for the PRM boundary was 11 mm ventral to the anterior aspect of the perineal body, and for the ICM boundary was 19 mm cranial to the arcuate pubic ligament (see Results for further details).

Measurement of maximum PVM cross-sectional area in the injury zone

The protocol to measure the maximum PVM cross-sectional in the injury zone using 3D Slicer (step 1-7) and Image J (step 8-9) was as follows:

1. Import and align the three standard orthogonal imaging planes (i.e., axial, sagittal and coronal)

2. Identify the PVM line-of-action (Figure 3 i) in a parasagittal plane.

3. Rotate the coronal plane perpendicular to PVM line-of-action (Figure 3 ii)

4. Identify the arcuate pubic ligament – move the coronal plane (as in step 2) posterior-anterior to find the first plane in which the ligament is seen (Figure 3 iii)

5. Draw a line that is 19 mm long perpendicular to the arcuate pubic ligament (Figure 3 iii)

6. In a sagittal view identify the anterior aspect of the perineal body (Figure 3 iv)

7. Place the coronal plane to show that same structure (anterior aspect of the perineal body) and move it 11 mm ventrally (Figure 3 v)

8. Outline the muscle bilaterally in every slice until the PVM arises from the pubic bone (include the slice defined in step 7) (Fig. 3 vi)

9. From the outlined PVM CSA, select the largest as the maximum PVM CSA

For intra-rater repeatability the measurements made by each examiner during two different sessions separated by approximately 2 weeks were used. For inter- rater repeatability, and also for reporting on the measurements, the first of these two sessions were used.

Statistical analysis

The reported anatomic CSA measurements concern the experts’ measurements at the first-time point (Table 1). To assess the mean cross-sectional area (PVM CSA), the measurements of both left and right CSA where considered for each subject, and the largest of the pair was selected for statistical analysis.

To find inter-and intra-rater reliability paired t-tests were used to assess differences between raters and left and right CSA. Intraclass Correlation Coefficients (ICC) (mode: two-way random; type: absolute agreement) (95% Confidence Interval) [14] were used to assess measurement agreement between raters and within a rater over time. All statistical analysis was conducted using SPSS (version 23, IBM, Chicago, IL, USA) and p <0.05 indicated statistical significance. All data are reported as mean and standard deviation (± SD) and percentiles were calculated for the area measurements.

Results

Based on the analysis of the women with unilateral tear and the women who had an elective C-section used for method development (Table 2), the PRM boundary was a mean 10.7 (± 5.1) mm ventral to the anterior aspect of the perineal body (median = 11 mm). The mean distance to the ICM boundary was 18.9 (± 3.5) mm cranial from the dorsal limit of the arcuate pubic ligament (median = 19 mm).

Table 2 –

Subject demographics by group: Unilateral tear and C-section, the later includes both elective C-section and C-section with 2nd stage of labor. No data was available (NA) for the duration of 2nd stage of labor for the women with a unilateral tear (Data are presented as mean (±SD), except for parity that is presented as median (IQR)).

Group	n	Age (years)	BMI (kg/m2)	Vaginal Deliveries	Duration 2nd Stage of Labor (min)	Time from birth to scanning (months)
Unilateral Tear	11	57.0 (± 7.9)	25.1 (± 4.1)	2 (1, 2)	NA	>60
C-section	24	28.9 (± 6.4)	27.0 (± 6.9)	0	228 (± 176)	8.1 (± 2.2)
Elective C-section	7	29.4 (± 5.1)	25.2 (± 5.0)	0	0	8.2 (± 1.9)
C-section with 2nd Stage	17	28.6 (± 6.9)	27.8 (± 7.4)	0	322 (± 116)	8.1 (± 2.3)

The average PVM line-of-action inclination angle to the horizontal reference plane corrected for variations in pelvic tilt [15] was 33° (± 5, range 21-43) in the 24 women who had undergone C-section (Table 2).

The mean maximum PVM CSA in the injury zone for the 24 subjects, all with intact PVM, was 1.25 cm² (± 0.29, range 0.75- 1.86, coefficient of variation 0.23). The 5th, 50th and 95th percentiles were also calculated as 0.77, 1.23 and 1.80 cm², respectively. No significant difference in maximum PVM CSA was found between left and right muscles (1.18±0.28 cm² vs 1.15±0.32 cm², p=0.46). The “belly” of the muscle, as assessed by the maximum unilateral PVM CSA was located a mean 7.3 mm (± 8.0, range 0.0-30.0) and 6.5 mm (± 6.8, range 0.0- 28.7) from the PRM boundary on the right and left sides, respectively.

In assessing inter-rater reliability, direct measurements of maximum PVM cross-section for each side were possible in all scans for both raters. Measurements between raters had an ICC of 0.95 (95% Cl 0.90-0.98, p=0.29) at the first time-point and ICC of 0.89 (95% CI 0.80-0.95, p=0.15) at the second time-point. The intra-rater assessment for each rater was also found to be similar comparing the two-measured time-points (ICC: 0.89, 95% CI 0.75-0.95, p=0.09) and (ICC: 0.90, 95% CI 0.79- 0.95, p=0.24). The difference between the PVM CSA measurements for each of the two raters was 6 ± 4%.

Discussion

This paper describes a new method for directly measuring maximum anatomic PVM CSA on MRI. It provides a way to quantify muscle bulk in the physiologically appropriate way: directly from the MRI using a plane that is normal to the recently described PVM line-of-action and muscle fiber direction [16] in the physiologically important thickest part of the muscle and avoids the artifacts introduced by the multi-step process of model creation and measurement. The maximum PVM CSA measured in the injury zone represents the maximum isometric force that the PVM generates based on its pubic bone insertion, whether the force transmission is normal or abnormal due to injury and compensated via lateral force transmission of adjacent structures.

Two difficulties arise in attempting to measure the CSA of the PVM. First, the PVM blends anatomically with the neighboring ICM without a well-defined separating fascia to signal the transition from the one to the other. Second, the PVM lies immediately adjacent to, and overlaps the PRM in its dorsal portion. Therefore, developing a method that allows a standard volume of PVM to be identified in which to seek its maximum CSA is needed and that was a goal of the present paper. Lack of such a method has led to the wide variations in PVM area [cf 11,17,18] and thickness [cf 8,19,20] measurements. In addition, specific anatomical landmarks that remain visible even in the absence of the PVM were used to identify its location where it does not overlap with the ICM or PRM. This is the first time that this separation has been made. Prior studies used planes of separation that were defined based on static anatomical landmarks, but not derived from actual measurements made on subjects. By using measured distances to define the PVM anatomical boundaries it is possible to identify its expected location even when it is missing. The acceptable intra-rater and inter-rater reliability for this analysis indicates that it is possible to achieve consistent results in measuring maximum PVM CSA.

The values for the maximum unilateral PVM CSA presented here are substantially smaller than published values based on lofted model measurement techniques (cf present value of 1.25 (±0.29) cm² vs 2.44 cm² [11] and 4.06 cm² [17]). We believe these differences result from different measurement techniques and protocols. Specifically, because the present area measurements were made orthogonal to the PVM line-of-action using a protocol that separates the PVM from the adjacent PRM and ICM, we believe the measurement artefacts to be smaller and inclusion of PRM mass within the measured area avoided: the location of maximum CSA, whether on the left or right side, was at least 6 mm ventral to the PRM boundary. We found the coefficient of variation in PVM CSA to be similar to that found in other small striated parallel-fibered muscles, such as the lumbar multifidus [for example, 19].

The recent description of MRI measurement techniques to determine the true PVM line-of-action [16] was fundamental to the accuracy of the present orthogonal measurements. In addition, this study has allowed us to recognize that the PVM CSA varies along its length, much as the biceps muscle does, with the CSA being much smaller near its pubic origin, presumably due to higher type 1 collagen content at the pubic aponeurosis [22]. The present method allows the maximal PVM CSA to be identified as one moves distally in the muscle and some images distinctly show that the muscle has a ‘belly’.

The definition of the PVM boundaries was based on three different concepts: the angle of its fiber orientation and hence its line-of-action, as well as the definition of its boundaries with the ICM and PRM. The measured PVM fiber direction angle was smaller than that reported by Betschart et al. (33 ± 5° vs 41 ± 8°)[16]. Our study, identified the direction of the striations within the muscle, while Betschart selected scans in which the muscle fibers were visible; a feature not consistent in all scans. A simple trigonometric calculation shows a difference of 8° would result in 1% change in the maximum measured unilateral PVM CSA and so it is likely not important.

The injury zone boundaries were calculated as 19 mm cranial from the dorsal limit of the arcuate pubic ligament for the ICM boundary and 11 mm ventral to the anterior aspect of the perineal body for the PRM boundary, and the maximum anatomic PVM CSA was measured within these limits. There is a risk that, in a woman who is smaller than those measured here, this might lead to inclusion of the small portions of the ICM and/or PRM in the measured PVM CSA; conversely, in a very large woman one might underestimate the PVM anatomic CSA. The magnitude of this potential effect can be addressed in the future when larger sample sizes make this type of analysis possible. Clearly, with a larger sample size, the present values for the 5^th and 95^th percentile CSA might be expected to be conservative, but the mean value should not change much.

The ability to make measurements directly from MRI rather than having to outline the muscle and make a 3D model of the PVM is a strength of the present measurement protocol. A limitation of this technique-development study is the small sample size and the fact that it is not population-based, and that measurement of CSA does not inform us about the function and quality of the outlined tissue (i.e., extent of fatty infiltration within the muscle) and how these characteristics change with age and PVM tear. However, neither the CSA [17] nor the maximum vaginal closure force [23] appear to be affected by age in nullipara suggesting that intact muscle quality does not significantly change with age.

In addition, the reliability studies were performed on images from the same subjects as those used for protocol development; however, these were the only subjects available with these unusual characteristics in our MRI database. We are not able to validate these measurements against a “gold standard”, because measuring the PVM CSA in cadaver specimens would not provide meaningful information for the living: the pelvic floor is too grossly distorted by loss of muscle tone after death [24]. Now that a technique for measuring muscle CSA is feasible and shows acceptable reproducibility, PVM CSA can be used to understand its interaction with the mechanisms of prolapse.

Brief Summary.

A technique is presented for measuring the anatomic cross-sectional area of the pubovisceral muscle directly on MRI of 24 primiparous women.