Estimation of Shear Wave Speed in the Rhesus Macaques Uterine Cervix

Abstract

Cervical softness is a critical parameter in pregnancy. Clinically, preterm birth is associated with premature cervical softening and post-dates birth is associated with delayed cervical softening. In practice, the assessment of softness is subjective, based on digital examination. Fortunately, objective, quantitative techniques to assess softness and other parameters associated with microstructural cervical change are emerging. One of these is shear wave speed (SWS) estimation. In principle, this allows objective characterization of stiffness because waves travel more slowly in softer tissue. We are studying SWS in humans and rhesus macaques, the latter in order to accelerate translation from bench to bedside. For the current study, we estimated SWS in ex vivo cervices of rhesus macaques, n=24 nulliparous (never given birth) and n=9 multiparous (delivered at least 1 baby). Misoprostol (a prostaglandin used to soften human cervices prior to gynecological procedures) was administered to 13 macaques prior to necropsy (nulliparous: 7, multiparous: 6). SWS measurements were made at predetermined locations from the distal to proximal end of the cervix on both the anterior and posterior cervix, with 5 repeat measures at each location. The intent was to explore macaque cervical microstructure, including biological and spatial variability, to elucidate the similarities and differences between the macaque and the human cervix in order to facilitate future in vivo studies. We found that SWS is dependent on location in the normal nonpregnant macaque cervix, as in the human cervix. Unlike the human cervix, we detected no difference between ripened and unripened rhesus macaque cervix samples, nor nulliparous versus multiparous samples, although we observed a trend toward stiffer tissue in nulliparous samples. We found rhesus macaque cervix to be much stiffer than human, which is important for technique refinement. These findings are useful for guiding study of cervical microstructure in both humans and macaques.

I. Introduction

The cervix undergoes extensive microstructural remodeling during pregnancy. This process is presently the subject of wide research effort because of its relevance to the problem of preterm birth, the leading global cause of neonatal death.¹ Cervical softening is one critical feature of this process. Softening commences soon after conception and gradually progresses, but the normal cervix remains stiff enough to maintain the fetus in utero during most of pregnancy. Near the end, an accelerated phase of softening (ripening) occurs, allowing the cervix to dilate completely for delivery of the fetus.^2;3 Clinical assessment of cervical softness is subjective, determined by digital evaluation. A quantitative, objective description of softness is imperative to comprehensive study of cervical remodeling and ripening in pregnancy in order to understand the molecular events and biomechanical forces that underlie cervical microstructural changes.

Shear wave elasticity imaging (SWEI) is one technique for objective assessment of tissue stiffness. Cervical microstructure is extremely complex, making this approach challenging, but we and others have demonstrated that it is feasible in humans.^4–8 Animal models are essential for accelerating research from bench to bedside. For pregnancy studies, sheep and rodents are the two most common models, primarily for logistical reasons, namely that sheep reproductive organs are large and easily accessible, and mice are inexpensive and have very short gestational periods that allows rapid study (21 days compared to 152 in sheep, 164 in rhesus macaques, and 280 in humans).⁹ These models have provided valuable insight into the structure and role of the cervix in preterm birth.^9–12 SWEI has been studied in vivo in a sheep model.¹³ Unfortunately, sheep and rodents are far from ideal for comprehensive study. One issue is that they have bicornuate (two) uteri and they are obligate quadripeds (they walk on four feet, which governs biomechanical forces on the cervix).^9;14 More importantly, unlike humans, labor is preceded by serum progesterone withdrawal, which is indicative of different mechanisms of parturition initiation.⁹ A further issue is that the placenta, the origin of hormones that control most pregnancy functions, is cotyledenous in sheep (as opposed to discoid in humans), which carries significant implications for control of mechanisms of cervical ripening and labor.⁹

In contrast, the pregnancies of nonhuman primates such as rhesus macaques are very similar to humans.¹⁵ Specifically, humans and nonhuman primates have similar maternal-fetal immunology and physiology, menstrual cycles (monthly), zygosity patterns (singleton fetus is most common), placentas (discoid), and parturition patterns (serum progesterone does not decrease prior to labor).^9;16–18 For these reasons, investigative paradigms involving pregnancy, embryo transfer, contraception and infection-mediated preterm birth are well established in the macaque.^14;19–24 Of particular relevance to human studies is that the macaque, unlike all other animal models used for studies of preterm birth, are not obligate quadripeds, which means that biomechanical forces on the cervix are likely to be more similar to humans. A well described feature of the macaque cervix is its serpentine canal,¹⁵ but to our knowledge, its microstructure has not been previously established.

Here we report our initial findings of SWEI exploration of relevant similarities and differences in the human versus macaque cervix. We used acoustic radiation force to generate shear waves in the cervical tissue and then measured shear wave speed (SWS) to objectively quantify tissue softness. This approach has been used successfully to estimate in vivo liver stiffness.^25;26 Our initial studies in the ex vivo nonpregnant human cervix demonstrated that this method provides low noise estimates of SWS and differentiates between ripened (with a prostaglandin agent, misoprostol, typically used for softening the human cervix prior to labor induction or gynecological procedures) and unripened tissue.^4;5 Mirroring our study in the ex vivo human cervix, we investigated spatial and biological variability in cervical tissue samples from nonpregnant nulliparous (never given birth) and multiparous (have given birth) rhesus macaques before and after administration of misoprostol. We investigated the entire cervix, but focused on the internal os. This is the most clinically relevant region in humans, where cervical change appears to initiate.²⁷ This clinical observation is supported by emerging data about cervical tissue extracellular matrix (ECM). The ECM contains interweaving zones of collagen that change along the cervix from the internal to external os,²⁸ serving as a scaffold and governing biochemical and mechanical tissue properties.^29;30 Tissue strength depends upon the degree and type of collagen crosslinks,^31;32 and recent studies show that the internal os has marked collagen crosslink heterogeneity compared to the external os.³³ Further, anatomically correct (for humans) computer simulation models have shown that the angle of the cervix in the pelvis influences the amount of stretch exerted on the area of the internal os,³⁴ as do the fetal membranes.³⁵ Taken together, these findings suggest that the function of the internal os differs from that of the external, with the internal os likely playing a more significant role in pregnancy. In light of these recent findings, it is unsurprising that our previous human studies using SWS demonstrated spatial variability along the length of the cervix.^4;5

We undertook this study because exploring the similarities and differences in humans and rhesus macaques in parameters such as cervical softening, especially at the internal os, is a necessary step toward facilitating development of a rhesus macaque model of cervical remodeling in pregnancy, which in turn should accelerate studies of preterm birth.

II. Methods

A. Tissue Acquisition and Preparation

Cervical specimens were obtained through the Nonhuman Primate Biological Materials Distribution Core at the Wisconsin National Primate Research Center. The entire cervix, vaginal cuff, and lower uterus was collected from 24 nulliparous and 9 multiparous adult rhesus macaques (n=33). Animals having previous cesarean section (associated with failed cervical ripening) or preterm birth (associated with premature ripening) were excluded. Seven nulliparous and 6 multiparous animals (n=13) assigned to primary projects not involving or affecting the reproductive organs were randomly selected and co-assigned to received 200 mcg misoprostol, administered orally or transvaginally, 12 hours prior to necropsy (see Table 1). The remaining samples (n=20) were untreated. All protocols were approved by the University of Wisconsin Institutional Animal Care and Use Committee.

TABLE I.

Summary of the number and status of the rhesus macaque cervix specimens.

	Multiparous	Nulliparous
Misoprostol	6	7
Unripened	3	17
TOTAL	9	24

Misoprostol is a prostaglandin used routinely in humans for cervical ripening prior to induction of labor or a variety of gynecologic procedures. Misoprostol is administered in combination with other pharmaceuticals in contraceptive studies in macaque species.^14;23 Further, there is evidence for a major role of prostaglandins in the initiation of parturition in rhesus, just like in humans; intramuscular or intravenous prostaglandins induce preterm labor and abortion in pregnant macaques.^36;37 Although there are no known established protocols for cervical ripening in the rhesus macaque because these animals do not typically undergo procedures (gynecologic or labor induction) that require ripening, in order to mirror our humans studies and establish whether SWS could distinguish between the ripened and unripened macaque cervix, we found it reasonable to use misoprostol. We used a dosing and administration protocol based on its use in nonpregnant, premenopausal women.^38;39

Reproductive tracts were dissected by a veterinary pathologist (HAS) who identified anatomic orientation. Specimens (length: 41.2±9.3 mm) were submerged in normal saline and brought to room temperature in order to reduce potential variability in data acquisition due to temperature. The sample was delicately pinned to a piece of sound absorbing rubber to prevent stretching or distortion of tissue, as well as minimize reverberation and motion artifact, analogous to our prior human study.^4;5

B. Data Acquisition and Processing

The experimental procedures, data acquisition and processing paralleled our previous study on the ex vivo human cervix.^4;5 Shear wave elasticity data were acquired using a modified Siemens Acuson S2000 ultrasound system with a 9L4 linear array transducer via the Virtual Touch Quantification software package (Siemens Healthcare, Ultrasound Business Unit, Mountain View, CA, USA). Imaging parameters are summarized in Table II. The 38 mm aperture array transducer was securely attached to a three dimensional positioning stage and aligned parallel to the endocervical canal. The focal depth was set at 3 cm (close to the elevational focus). The experimental set-up is the same as shown in Figure 1 of Carlson et al.⁵ Following ultrasound data acquisition, we performed microscopy on the imaged tissue in order to corroborate our SWS data.

TABLE II.

Summary of imaging system parameters for SWS data acquisition. The f-number is defined as the ratio of focal depth to active aperture size. Track PRF depends on details of the experimental configuration (e.g., depth of imaging), but PRF values were generally about 9kHz.

Parameter	9L4 Linear Array
Push frequency (MHz)	4
Track frequency (MHz)	6.16
Push focus (mm)	30
Elevation focus (mm)	48
Pulse cycles (n)	400
Pulse duration (μs)	100
Track pulse repetition frequency (kHz)	~9
Push f-number	1.5
Track f-number	1.5
Mechanical index (push pulse)	1.9

Fig. 1.

B-mode image and low magnification light microscopy images of cervix specimens. The B-mode image was acquired with the transducer aligned parallel to the endocervical canal and focused mid-depth through the cervix. The coupling medium was saline. Shear wave data were acquired at ROIs along the length of the cervix. The microscopy images were acquired from 200 μm thick slices of two rhesus macaque cervical samples (top image 5.2 years old; lower image 10.9 years old). The B-mode image was acquired from the approximate dash-lined region of the top cervical sample.

Figure 1a shows the B-mode image of a cervix specimen. Measurements were made along the length of the cervix between the distal (external os; the vaginal end) and proximal (internal os; the uterine end) cervix at carefully selected locations with 5×5 mm, 50% overlapping, regions of interest (ROI). Five measurements were obtained at each location to improve estimate accuracy. ROIs were centered midway through the thickness of the cervix (from canal to outer edge) and carefully chosen to avoid regions of heterogeneity on B-mode images to avoid any unpredictable behavior caused by boundaries. After one side was scanned, the specimen was flipped and the same procedures were repeated on the other side to acquire data on both anterior and posterior sides.

C. Data Analysis

After data acquisition, MATLAB (Mathworks, Natick, MA, USA) was used for offline data analysis. Tissue displacement was estimated using Loupas’ two-dimensional autocorrelator on in-phase and quadrature (IQ) data.⁴⁰ Five repeat displacement estimates from the same ROI were averaged, and then filtered using a 0.98 correlation coefficient threshold to remove poor displacement estimates. SWS (group velocity) was estimated from filtered displacements by the random sample consensus (RANSAC) algorithm. Specifically, the time-to-peak (TTP) values were extracted from displacements to fit a RANSAC paradigm that minimized the number of outliers and sum of the square errors of the inliers. SWS was then estimated from the inverse slope of the fitting plane in lateral direction.⁴¹ After estimation, criteria for inclusion in further data analysis were based on a minimum peak displacement caused by the shear wave (at least 0.5 μm) and minimum fraction of the displacement estimates contributing to the surface fit⁴¹ (inliers must exceed 40%). SWS not meeting these criteria were excluded from further analysis.

The rhesus macaque cervix presented some interesting challenges compared to the human cervix. We found the macaque cervix is shorter with greater individual variation than the human cervix. Low magnification microscopy images (PathScan Enabler IV, Meyer Instruments, Houston, TX) of 200 μm thick slices extracted from each cervix specimen, such as those in Fig. 1b, demonstrate some of the differences. The human cervix is generally cylindrical, with a relatively straight cervical canal from its proximal to distal end. In contrast, the canal of the rhesus macaque is generally serpentine,¹⁵ particularly distally. We noticed that, in fact, in some cases the canal was serpentine throughout. Also, some rhesus macaque cervices were nearly spherical and lacked any apparent homology with the human cervix (lower image in Fig. 1b). These findings were important for two reasons. First, significant microstructural hetereogeneity and complexity violates the assumption of tissue homogeneity for SWS estimation. Second, measurements from tissue lacking any similarity to human cervix are unlikely to be meaningful to our exploration. As with our previous human study, SWS estimates from cervices of different lengths were compared via division of the cervix into 5 parts (distal, mid-distal, middle, mid-proximal and proximal) according to its length. In this study, the distal and mid-distal end of all specimens, and the middle region of many specimens, were excluded from analysis because of lack of homology with human cervix. In most cases, this did not markedly affect our study because our primary area of interest is the internal os (proximal region). However, in 3 (of 33) samples, we suspected an entirely spherical cervix via B-mode ultrasound and, after this was confirmed via microscopy, these were excluded from analysis due to lack of relevance to human studies.

The Kolmogorov-Smirnov test was used as a means to assess the statistical significance of differences between (1) variability of SWS in the anterior vs posterior cervix, (2) spatial variability of SWS averaged over repeat measures as a function of position along the length of the cervix, (3) sensitivity of SWS to distinguish misoprostol-ripened from unripened (control) cervical tissue, and (4) sensitivity of SWS to distinguish nulliparous from multiparous cervical tissue. A probability of equal means less than 0.05 (two-sided) was used as the criterion for statistical significance.

III. Results

A. Spatial Variability of SWS

Figure 2 shows estimates of SWS demonstrating spatial variability from the middle to the proximal region of the cervix and in its anterior versus posterior segments. All groups display a spatial slope in SWS, indicating a decrease in SWS from the middle to proximal end. A least-squares fit to the mean SWS at each position (middle to proximal) was computed to estimate the spatial gradient. SWS gradients in the anterior sections were −1.96 and −2.44 m·s⁻¹cm⁻¹ for the unripened ripened and groups, respectively. The corresponding gradients in the posterior group sections were −2.44 and −1.97 m·s⁻¹cm⁻¹ for unripened and ripened groups, respectively (using the average mid-to-proximal cervix length of 1.0 cm among our samples). The differences were not significant. We determined whether anterior and posterior groups could be combined into a single group in order to increase the statistical power of comparisons between ripened and unripened tissue. As expected, these groups were not significantly different. The p-values comparing anterior and posterior specimens for the unripened group were 0.874, 0.356, and 0.095 for middle, mid-proximal, and proximal locations, respectively. For the ripened group, the p-values between anterior and posterior specimen were 0.734, 0.742, and 0.130 for middle, mid-proximal, and proximal locations, respectively. Because there were no statistically significant differences, the anterior and posterior data were combined for all remaining comparisons.

Fig. 2.

Box plots of mean SWS from each specimen according to location in the cervix for ripened and unripened specimens. On each box, the central mark is the median value of the SWS estimates, with the edges of the box representing 25th and 75th percentiles (the interquartile range; IQR), the whiskers display the maxima and minima at each location within 1.5×IQR. The notch shows the 95% confidence interval for the median value. Outliers beyond 1.5×IQR are removed (the number of outliers in (a), (b), (c), and (d) are 3, 3, 2 and 1, respectively). The numbers under each box indicate the mean SWS±standard deviation and the total number of specimens for each box (in parentheses).

B. Misoprostol-ripened versus unripened cervix

Figure 3 shows box plots comparing SWS estimates for ripened and unripened cervical tissue. The SWS (mean ± standard deviation) for the unripened group were 5.80±1.95 m/s, and 5.85±1.97 m/s for the ripened group in the middle location. There was no significant difference in SWS estimates between unripened and ripened tissue. For example, the p-values were 0.944, 0.058, and 0.179 for the middle, mid-proximal, and proximal locations, respectively. Further, larger standard deviations were seen among ripened specimens compared to unripened specimens in all 3 spatial locations.

Fig. 3.

Box plot of mean SWS for each spatial location combining anterior and posterior measurements in the ripened or unripened groups. Outliers beyond 1.5×IQR are removed (the number of outliers in (a), (b), and (c) are 2, 5 and 1, respectively).

C. Nulliparous versus Multiparous Cervix

Figure 4 shows the comparison of SWS estimates between nulliparous and multiparous cervical tissues. We found a trend toward different SWS in the nulliparous group compared to the multiparous group. At the middle location, the mean SWS were 6.09±2.09 m/s and 4.97±1.27 m/s for the nulliparous unripened group and multiparous unripened group, respectively. The mean SWS were 6.09±2.09 m/s and 5.59±1.87 m/s for the nulliparous ripened group and multiparous ripened group, respectively. Although the sample sizes are low (multiparous n=2 and n=5 for the unripened and ripened group respectively), the results suggest a nonsignificant trend toward softer tissue in the multiparous cervix.

Fig. 4.

Box plot of mean SWS from each sample according to parity (nulliparous vs. multiparous) at different locations. Outliers beyond 1.5×IQR are removed (the number of outliers in (a), (b), (c), (d), (e) and (f) are 1, 1, 4, 0, 4 and 1, respectively).

IV. Discussion

Shear wave elasticity imaging has the potential to advance studies of cervical microstructure. We and others have demonstrated its feasibility in humans and animal models.^4;6–8;13 This work represents our initial exploration of SWS estimation in the rhesus macaque model we are developing for comprehensive study of the pregnant cervix. The rhesus macaque carries several advantages with regard to homology with humans. However, this study elucidated some important differences that are relevant to SWS estimation. For instance, the macaque cervix is even more heterogeneous than the human cervix, with areas containing whorls of collagen that create extremely complicated boundary conditions. This is relevant because one limitation of SWEI in the cervix is that, for SWS estimation, the tissue under evaluation is assumed to be isotropic, homogeneous and semi-infinite in the region where the wave speed is measured^42–44 but in fact cervical microstructure is anisotropic and heterogeneous as discussed above and shown in Fig. 1b. Despite this limitation, we have previously demonstrated SWS estimation feasibility under experimental conditions similar to those used in this study in both the ex vivo human cervix^4;5 and simulations.⁴⁵ In humans, optimal measurement feasibility and reliability were found in the proximal cervix, and this study confirms that finding in macaques as well. We presume that is because proximal cervical microstructure in both species is quasi-homogeneous, thus the assumptions of SWS estimation are not violated. Fortunately, it is the proximal cervix that is best for study in both humans and macaques, allowing synergy in future study design. Also fortunate for in vivo study is that these complex areas are avoidable because they are obvious upon B-mode imaging (in fact, that observation is what led us to microscopy studies to investigate this finding we had not observed in humans).

Another important finding is that the macaque cervix is much stiffer than the human, as evidenced by SWS estimates that are about 2-fold higher; the average SWS from a mid-proximal location of the macaque cervix was ~5.8 m/s compared to ~3.5 m/s from the equivalent location in the human cervix. This is relevant because SWS estimation is less precise in stiffer materials.^46–48 Specifically, displacements from stiffer materials have lower amplitudes and are more easily corrupted by noise, with the uncertainty of the SWS estimation linearly proportional to the square of the SWS.^47;48 This may explain why we found standard deviations in macaque cervix SWS estimates to be larger than those in ex vivo human cervix estimates (>1 m/s versus <1 m/s, respectively).^4;5 A related issue is that, for human studies, it is reasonable to assume a constant SWS throughout a 5×5 mm ROI because we have shown in the ex vivo human cervix that shear wave acceleration due to its velocity gradient within this ROI has minimal effect on SWS estimates.⁵ However, in the rhesus macaque cervix the gradient in SWS is about twice as large and therefore variability in ROI positioning by half its width (1/2 of 5 mm) could lead to a SWS estimate bias of as much as about 0.5 m/s (a modest component of the overall uncertainty in SWS estimates for any particular specimen).

Unlike human studies, we did not see a difference in ripened compared to unripened cervical tissue, and, although we did find marked variability within and between macaque cervices, as in humans, the stiffness trend along the canal was opposite, with the macaque proximal cervix softer than distal areas.⁴ While these differences must be kept in mind, they do not preclude meaningful study. Specifically, we expected to see a difference in SWS estimates in ripened compared to unripened cervical tissue in this study given known events related to cervical ripening in the human and mouse, namely, increased hydration and loss of collagen crosslinking.^2;31;49;50 Concordant with this, significant differences in SWS estimates between unripened and ripened tissue have been found in vivo and ex vivo in humans^4;6 and sheep.¹³ While higher estimate uncertainty in the stiffer macaque cervix could mask small differences in stiffness, a more likely reason is that the misoprostol protocol we used is ineffective in the nonpregnant macaque. In contrast, the efficacy of misoprostol for cervical ripening in both pregnant and non-pregnant humans is well established, as is the efficacy of dexamethasone, the agent used in the study of pregnant ewes.¹³ As discussed in the Methods, despite the absence of published reports of misoprostol alone for ripening the macaque cervix, particularly the nonpregnant macaque cervix, we chose this agent in order to harmonize this study with our human studies. This approached seemed reasonable because, it is known that macaques have decreased sensitivity to certain prostaglandins,^22;51 in pregnant macaques, intramuscular or intravenous prostaglandins induce preterm labor and abortion,^36;37 misoprostol is used in combination with other agents in contraceptive studies^14;23 and to induce labor in post-term pregnancy or promote expulsion of retained products of conception,^22;51;52 and intracervical application of a different prostaglandin (PGE2 gel) causes cervical ripening and decreased collagen fibers.⁵³ However, we observed no difference in ripened versus unripened macaque tissue despite consistent SWS estimate trends that gave us confidence in our measurements. We therefore concluded that the intervention was ineffective and unnecessarily invasive, and therefore abandoned it. We did not explore different misoprostol doses or routes of administration, or other pharmaceuticals, because the point of this study was to explore similarities and differences between macaques and humans, not to determine how to ripen the macaque cervix.

Another interesting finding was that stiffness trends in the macaque cervix were opposite those in the human. Again, because trends were consistent, we trust our measurements. While we did not uncover previous reports of this finding, our veterinary pathologist (HAS) believes the stiffness trend (and distal whorls) in the macaque cervix is likely teleological. Specifically, there are striking gait differences between the two species, with macaques exhibiting regular quadripedal to bipedal locomotion compared to exclusively bipedal humans. In addition macaques have leaping and arboreal (tree climbing) habits developed to avoid predators.⁵⁴ These factors logically dictate a more distally robust cervical structure during pregnancy.

Regarding the trend toward stiffer tissue in nulliparous versus multiparous macaques, this is interesting because the concept agrees with the (subjective) opinion among clinicians that the (pregnant or nonpregnant) nulliparous cervix is stiffer than the multiparous. Because hysterectomy in nulliparous women is relatively uncommon, numbers were too low in our human study (3 nulliparous specimens among more than 40 total) for adequate comparison. A recent study in pregnant ewes⁸ also noted no stiffness differences in (in vivo) nulliparous cervical tissue compared to multiparous, although this study included 9 sheep total, only 2 of which were nulliparous (one in the ripening group and one in the no intervention group), which makes it difficult to draw conclusions. It is worth noting that, in our study, the nulliparous macaque cervix may be stiffer, causing larger SWS estimate variability, but the number of multiparous specimens smaller, thus the estimate of the mean SWS in that group might be inaccurate. Whether the average nulliparous cervix (especially pregnant) differs from the multiparous is relevant to our studies of preterm birth, and we are currently investigating this in vivo in pregnant macaques.

Another consideration that further complicates analysis is that shear wave propagation is affected by tissue viscosity which leads to shear wave speed dispersion. It is reasonable to think that viscosity is changing in the cervix during pregnancy since it has been demonstrated that acoustic attenuation varies during pregnancy.⁵⁵ In humans and rodents, it is presumed that viscoelasticity increases as the cervix softens, and especially during ripening, because collagen crosslinking decreases with concomitant increase in the ratio of immature to mature collagens, and because of high molecular weight hyaluronin which causes increased hydration.^2;31;49;56 These factors might lead to changes in loss mechanisms for shear waves. Our data interpretation did not account for shear wave viscosity or dispersion, which is important because the estimated group velocity of shear waves (the quantity reported here and on commercial ultrasound imaging systems) is dependent on the the amount of dispersion and frequency content of the wave.^45;57 The frequency content of the shear waves in cervix is considerably higher than that of liver,⁵⁸ for example, and the frequency content of shear waves varied among specimens, making these measurements more sensitive to dispersion than those in liver, and possibly more variable among specimens. Preliminary data⁵⁹ suggest that human and rhesus macaque cervix are dispersive, which suggests an additional component of variance in our data that deserves further investigation.

One caution is that the SWS estimates we report are unlikely to be representative of in vivo results because experiments were performed at room (not body) temperature, ex vivo tissue lacks turgor as compared to in vivo tissue, and we used a transducer that is not appropriate for in vivo intracavity scanning. Specifically, as we transition to in vivo study, we remain mindful that our intent was to explore spatial variability in the macaque cervix, as well as similarities and differences between the macaque and human cervix, and not to establish absolute stiffness values that could be meaningfully translated to the in vivo situation.

In summary, this study successfully extended SWS estimation in the cervix to the rhesus macaque model and also provided an understanding of the similarities and differences between the macaque and human cervix. Specifically, we learned there is significant homology between macaques and humans in the proximal area of the cervix. This is both logical and reassuring given increasing clinical and basic science research that suggest that targeting this area is likely to be most productive in terms of understanding abnormal cervical remodeling. Also, this work establishes that there is significant variability within and among rhesus macaque cervices, as is the case with human cervices. Whether the average nulliparous cervix differs in stiffness from the average multiparous is clinically relevant, and therefore the object of future study. For that and other reasons, we have longitudinal studies in pregnant macaques and humans, as well as parallel studies of shear wave speed dispersion, currently underway.

V. Conclusions

Shear wave elasticity imaging in the rhesus macaque cervix is challenging, and, given that higher SWS is associated with greater estimate variance, our results must be interpreted in light of the finding that the nonpregnant rhesus macaque cervix is much stiffer than the human. Despite this, SWS estimation is promising for assessing cervical softness in rhesus macaques, making this a strong animal model to accelerate studies of preterm birth.