Quantitative Ultrasound-Based Characterization of Placental Tissue Microstructure in a Rat Model of Preeclampsia

Abstract

Objective:

Quantitative ultrasound (QUS) methods were applied to detect changes in placental microstructure in the reduced uterine perfusion pressure (RUPP) model of preeclampsia in rats. Preeclampsia is a life-threatening pregnancy disorder associated with abnormal placental development which is inadequately treated and managed.

Methods:

Nine timed-pregnant Sprague-Dawley rats were used in this study. On gestational day 14 (GD14), five rats received the RUPP surgery to induce preeclampsia. Ultrasound radiofrequency data were acquired for QUS analysis on GD18. On GD19, animals were sacrificed and dissected to acquire placental tissue samples. The cell nuclear diameter in each anatomical layer of the placenta was measured to compare with regional effective scatterer diameter (ESD) values.

Results:

ESD measurements obtained using in vivo QUS imaging correlated well (R² = 0.58, p = 3.8e-6) with cell nucleus diameter measurements from microscopy images. RUPP placentas had significantly smaller junctional zones compared to control placentas (p = 0.013). The average ESD in RUPP placentas was 1.0 μm smaller compared to control placentas (p = 0.040). This decrease in ESD in RUPP placentas is consistent with the decreased size of the junctional zone, which, in comparison to the labyrinth zone and chorionic plate, has larger cell nuclei (p = 3.3e-21 and p = 9.5e-27, respectively) and larger ESD (p = 5.6e-4 and p = 4.5e-8, respectively).

Conclusions:

These results demonstrate the potential of QUS as a noninvasive tool for detecting critical changes in placental microstructure, improving maternal and fetal outcomes by enabling earlier diagnosis and more timely therapeutic interventions in preeclampsia.

Introduction

Preeclampsia is a life-threatening pregnancy disorder that occurs when the placental microstructure fails to develop properly (1). The placenta is a transient organ that serves as the interface of maternal and fetal vasculature during pregnancy, delivering nutrients to and removing waste from the fetal circulation (2, 3). In normal pregnancies, the spiral arteries—which deliver nutrient-rich maternal blood to the placenta—remodel to increase in diameter, thereby increasing their capacity to meet the growing demands of the developing fetus. In preeclamptic pregnancies, the spiral arteries fail to remodel sufficiently, restricting growth of the fetus and triggering cardiovascular dysfunction in the mother (4). In severe cases, preeclampsia can evolve to eclampsia, causing the mother to experience multi-organ failure and seizures (5). Monitoring placental microstructure during pregnancy could help physicians to identify preeclampsia earlier and implement timely treatment.

Ultrasound (US) is a ubiquitous tool for non-invasive imaging during pregnancy because it is non-ionizing, real-time, portable and inexpensive compared to other medical imaging modalities such as magnetic resonance imaging or computed tomography (6); however, alterations in tissue microstructure do not necessarily cause visible morphological changes in conventional B-mode US images. In conventional B-mode imaging, US echoes are electronically recorded as radiofrequency (RF) data and converted into image pixels by digitizing the log-compressed RF envelope (7). In the envelope digitization process, valuable frequency information that can be analyzed to quantify tissue microstructural properties is discarded. By discarding frequency information, current clinical B-mode US images do not have sufficient resolution or contrast to resolve the microstructural elements in the placenta to detect the changes occurring in preeclampsia. Moreover, clinical B-mode image quality depends on the US image acquisition system as well as the user operating the imaging system, meaning that system-to-system and user-to-user variation can impact physician diagnosis (8).

Unlike conventional B-mode ultrasound, quantitative ultrasound (QUS) methods leverage the RF data to quantitatively assess tissue microstructure independent of imaging system and operator variability (9, 10). In this study, the RF data from US image acquisitions are used to compute the backscatter coefficient (BSC) and envelope statistics, which are then fit with theoretical models to obtain QUS parameters that describe the size, concentration, and organization of acoustic scatterers within the tissue of interest. Therefore, tissue scattering properties quantified using QUS methods likely correlate with tissue microstructure. QUS methods have been applied to assess changes in tissue microstructure in the placenta (11) as well as other organ systems, including liver (12, 13), breast (14), lymph node (15, 16), thyroid (17), eye (18), cervix (19, 20), and prostate (21, 22). Recently, QUS was applied to differentiate between normal and preeclamptic human placentas (11); however, this approach involved imaging placenta tissue ex vivo. Our goal is to noninvasively assess preeclampsia in vivo. Furthermore, this ex vivo approach did not account for the microstructural differences between the anatomical layers of the placenta, which may allow for more nuanced monitoring and assessment of preeclampsia.

The anatomical layers of the placenta have different microstructural properties to support different functions (23). Maternal blood is delivered to the placenta via the spiral arteries at the basal plate near where the placenta interfaces with the uterus. On the opposite side of the placenta, fetal blood is delivered to the placenta via the umbilical cord, which inserts into the placenta at the chorionic plate. Maternal and fetal blood supplies interact in the middle of the placenta, known as the labyrinth. In the labyrinth, fetal vasculature branches out from the chorionic plate into a pool of maternal blood to exchange nutrients and waste. Here, maternal and fetal blood supplies are separated by a multinucleated trophoblast cell layer known as the syncytiotrophoblast. The syncytiotrophoblast serves as a barrier from pathogens and a medium for nutrient and waste transport. Each layer of the placenta features a different microstructural composition to support the different functions that it serves (23).

In the present study, we investigate QUS methods to noninvasively detect differences in regional placental microstructure in an established rat model for preeclampsia. The reduced uterine perfusion pressure (RUPP) surgical model of preeclampsia reduces blood flow to the placenta, invoking the symptoms of hypertension and proteinuria observed in the human disease (24). Previous publications studying the RUPP model have reported that the region of the placenta which interfaces with the uterus—known as the junctional zone or basal zone in rats—is significantly smaller in RUPP placentas compared to control placentas (25, 26). These studies also found that the size of the junctional zone relative to the labyrinth zone (where nutrient exchange between mother and fetus occurs) is also smaller in RUPP placentas. Although the junctional zone and labyrinth zones have different microstructural compositions, there exists no contrast between these two regions in B-mode images. We hypothesized that QUS methods would provide contrast between the different anatomical regions of the placenta. We also hypothesized that QUS methods would be sensitive to preeclampsia because of the change in relative size of the junctional zone in the RUPP placentas. While the RUPP model in rats does not exhibit the same microstructural changes observed in human preeclampsia (syncytial knots, villous hyperplasia, etc. (27)), this model is sufficient to demonstrate the feasibility of in vivo regional placenta microstructure assessment in preeclampsia.

Our objective is to apply QUS to noninvasively study the microstructure of the placenta and assess the differences between normal and preeclamptic development in the RUPP rat model of preeclampsia in vivo.

Materials and Methods

Animal Model

This study was conducted following Protocol 1586 approved by Tulane University Institutional Animal Care and Use Committee (IACUC). Timed-pregnant female Sprague-Dawley rats (Charles River Laboratories, Boston, MA) were kept at Tulane’s animal facilities accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC).

On gestational day 14 (GD14), each rat underwent a surgical procedure: either the RUPP surgery to induce preeclampsia or a sham surgery to serve as a control. To minimize potential variability related to the surgical procedures, all surgeries were performed by a single researcher with more than two years of experience in the technique (28). The RUPP procedure involved the application of silver clamps to the uterine arteries and abdominal aorta to induce placental ischemia (29). The sham procedure involved opening the abdomen and externalizing the uterus, then returning the uterus and closing the abdomen. All surgical procedures were performed under anesthesia using 2–3% isoflurane mixed with oxygen gas administered via inhalation. For more detailed information on animal care for this study, please refer to the Supplementary Materials.

In vivo Imaging

US image data were acquired for QUS analysis (Fig. 1) on GD18. US RF data were collected using a Vevo 2100 equipped with the LZ250 transducer with 21 MHz center frequency (FUJIFILM VisualSonics, Toronto, Canada). We chose a 21 MHz center frequency transducer as it provides an optimal tradeoff between penetration depth and spatial resolution for abdominal imaging in small animals such as rats. Prior to imaging, animals were anesthetized with isoflurane as previously described. In preparation for imaging, the skin in the abdominal imaging region was shaved using clippers, depilated using Nair (Church & Dwight Co., Inc., Ewing, NJ), and coated with US gel to ensure coupling with minimal artifacts.

Figure 1:

Flowchart for generating effective scatterer diameter (ESD) parameter maps.

The abdomen was then scanned until a placenta was located. The angle of the imaging plane was then adjusted so that the insertion of the umbilical cord into the placenta was visible. Since the placenta is approximately cylindrically symmetric about the umbilical cord, we adjusted the imaging plane to a cross-section of the umbilical cord insertion to ensure consistency between placenta images. The presence of the umbilical cord was then confirmed using Doppler ultrasound imaging to identify the presence of blood flow.

RF data were recorded for 40 sequential images at 27 frames per second of 4–6 different placentas per animal. Image sequences with visibly detectable motion from maternal respiration or significant fetal movement which displaced the placenta were excluded from the study to avoid artifacts from blurring or skipping frames with respiration gating. This resulted in a total of 18 control placentas and 20 RUPP placentas for image and QUS analyses. For more detailed information about sample sizes, see below under Materials and Methods section Sample Size Determination.

QUS Processing

For each image sequence, specific anatomical features were manually segmented to guide the QUS image generation. First, the surface of the skin (Fig. 1) was traced to define where the bulk of tissue began. This helped align our imaging sample with images of our reference phantom for calibration. Second, we traced the border of the placenta (Fig. 1). This was to define the area for our QUS analyses of the placenta.

QUS parameter maps were generated for each placenta image sequence. Each pixel in a QUS parameter map corresponds to the value of a QUS parameter derived from an acoustic-scattering or envelope-statistics method applied to RF echo data in a 1 mm deep x 1.75 mm wide region of interest (ROI), about 10 times the wavelength at the center frequency axially and about 10 times the beam width laterally (30). The QUS ROI was moved across the segmented placenta to generate QUS parameter values throughout the field of view (Fig. 1). The QUS ROIs overlapped 75% in both the width and depth dimensions to increase the resolution of the QUS parameter maps with limited redundancy. QUS methods based on measurements of the BSC and envelope statistics have been previously described (10) and are briefly explained below.

QUS Parameter Computation

A Gaussian scattering model fit to the BSC provides information about the size and acoustic concentration of the scatterers (31). First, we computed the BSC within each ROI using the RF data from the placenta and RF data from a reference phantom with known scattering properties (32). BSC calculations were carried out using a processing bandwidth of 9 MHz to 29 MHz. The reference phantom consisted of 18-μm-diameter borosilicate beads suspended in an agar-based material with a 52-μm-thick TPX surface membrane (polymethylpentene, Mitsui Chemicals, Rye Brook, NY, USA).

The BSC for each region within the placenta was calculated as such:

$${BSC}_{ROI}\left(f,z\right)=\frac{{BSC}_{ref}\left(f\right)·NPS\left(f,z\right)·A\left(f,z\right)}{T\left(f,z\right)},$$ (1)

where $f$ is the frequency, $z$ is the distance from the skin surface to the center of the ROI, $NPS$ is the normalized power spectrum within the ROI described in (2), $A$ is the attenuation compensation described in (3), $T$ is the transmission coefficient through the TPX membrane of the reference phantom, and ${\mathrm{B}\mathrm{S}\mathrm{C}}_{\mathrm{r}\mathrm{e}\mathrm{f}}$ is the (known) BSC of the reference phantom.

The normalized power spectrum was defined as:

$$NPS\left(f,z\right)=\frac{1}{N·{PS}_{ref}\left(f,z\right)}{\sum }_{i=1}^N{PS}_{ROI}^i\left(f,z\right)$$ (2)

Where $\mathrm{N}$ is the number of RF line segments in the ROI, ${PS}_{ref}$ is the average power spectrum of the reference phantom, and ${PS}_{ROI}^i$ is the power spectrum of the i^th RF segment within the ROI. The reference phantom power spectra were calculated using the same depth, lateral extent, and axial range as the tissue data ROI.

Attenuation compensation was calculated as:

$$A\left(f,z\right)=exp\left(4\left(\alpha \left(f,z\right)\right)-{\alpha }_{ref}\left(f,z\right)\right)$$ (3)

For the placenta we defined $\alpha \left(f,z\right)=z·0.5f\mathrm{d}\mathrm{B}/\mathrm{M}\mathrm{H}\mathrm{z}/\mathrm{c}\mathrm{m}$, which is typical for soft tissues. For the reference phantom we defined ${\alpha }_{ref}(z,f)=z·{0.0279f}^{1.862}\mathrm{d}\mathrm{B}/{\mathrm{M}\mathrm{H}\mathrm{z}}^{1.862}/\mathrm{c}\mathrm{m}$. The transmission of the sound through the TPX membrane $T$ was calculated according (33) using speed of sound of 2093 m/s and acoustic impedance of 1.82 MRayl. The attenuation of the sound wave through the TPX membrane was modeled as $\alpha \left(f,z\right)=z·0.019f^{1.862}\mathrm{d}\mathrm{B}/{\mathrm{M}\mathrm{H}\mathrm{z}}^{1.862}/\mathrm{c}\mathrm{m}$. REF

${\mathrm{B}\mathrm{S}\mathrm{C}}_{\mathrm{r}\mathrm{e}\mathrm{f}}$ was calculated according to (34) with the following borosilicate bead properties: speed of sound = 5640 m/s, Poisson ratio = 0.2, and mass density = 2.5 g/cm³. A scattering model with a gaussian form factor was then fit to BSC according to the following equation (31):

$${BSC}_{GS}\left(f\right)=\frac{n{\gamma }_0^2{\pi }^4{f}^4{V}_s{\left(a_{eff}\right)}^2}{4{\pi }^2{c}^4}exp\left(\frac{-0.8274{\pi }^2{f}^2{a}_{eff}^2}{c^2}\right)$$ (4)

Where $n$ is the number density of scatterers, ${\gamma }_0$ is the relative impedance mismatch per scatterer, $V_s$ is the effective scatterer volume equal to (4/3) $\pi a_{eff}^3$, and $a_{eff}$ is the effective scatterer radius. From fitting this scattering model, we obtain the effective scatterer diameter (ESD), equal to twice the radius ${2a}_{eff}$.

Whole Placenta QUS Analysis

Our main findings correlate placental microstructure with ESD parameter maps. The first analysis looked at the average value of each ESD parameter map within each placenta (Fig. 2). For each parameter map, the average ESD parameter value within the placenta across 40 image acquisitions was computed to represent a given placenta. The average value of each placenta was then displayed on a violin plot, separating the RUPP and control groups. An independent two-sample t-test was used to calculate whether there was a statistically significant difference between the RUPP and control groups.

Figure 2:

Image processing workflow for QUS analyses of the rat placenta.

Regional Placenta QUS Analysis

A second analysis looked at how ESD parameter values changed across the different layers in each placenta (Fig. 2). Since it is not possible to distinguish between the different layers of the placenta in B-mode images, as it is in histology images, we used anatomical landmarks present in the B-mode images to approximate the locations of each structural layer.

To approximate the location of the junctional zone, a line was manually drawn to trace the convex edge of the placenta interfacing with the maternal tissue. While the thickness of the junctional zone varies between placentas, the location of the junctional zone always begins at the utero-placental interface and extends into the placenta. To ensure that the area sampled was indeed the junctional zone, we sampled the pixels within the placenta closest to the utero-placental interface. To ensure consistent spacing between our sampling points, we fit an ellipse to the traced edge and placed sampling points along the ellipse such that the angular distance between them was equal. Twenty sample points were placed along the ellipse arc which spanned the distance of the traced curve. Each of the 20 points sampled the QUS resolution cell within the placenta whose center was nearest to the ellipse sampling point. The average of 40 sequential image acquisitions across all 20 sampling points served to represent the average QUS parameter value for the junctional zone (Fig. 2).

A similar ellipse fitting technique was used to approximate the chorionic plate. Since the umbilical cord always inserts into the placenta at the chorionic plate, we used the umbilical cord’s location in the B-mode image to guide our approximation of the chorionic plate’s location. A line was manually drawn to trace the concave edge of the placenta surrounding the umbilical cord insertion, which was verified using Doppler ultrasound. The concave trace was also fitted with an ellipse arc with 20 angularly spaced points which each sampled the nearest QUS resolution cell within the placenta, averaging 40 sequential image acquisitions.

Since the labyrinth zone lies between the junctional zone and chorionic plate, we approximated its location by interpolating a new ellipse arc at the midpoint between the junctional zone and chorionic plate arcs. Since the labyrinth zone accounts for the bulk of the placenta’s volume, it is expected that this midpoint ellipse sampled points in the labyrinth zone. The labyrinth zone arc also included 20 angularly spaced points sampled within the placenta, averaged across 40 frames.

A two-way analysis of variance (ANOVA) was used to calculate the effect of the placenta layer and surgical group on the average parameter value. Once the variables with significant effect on variance were identified, an independent two-sample t-test was used to calculate whether there was a statistically significant difference.

Ex vivo Histological Analysis

On GD19, animals were sacrificed and dissected to acquire placental tissue samples. Three placentas were collected for histological analysis from each animal, resulting in 12 control placentas and 15 RUPP placentas. For more detailed information about sample sizes, see below under Materials and Methods section Sample Size Determination.

Each placenta was cut in half along the insertion of the umbilical cord, fixed in formalin, and embedded in paraffin for tissue sectioning. Tissues were cut into 4-μm-thick slices and stained with hematoxylin and eosin (H&E). Tissue sections were mounted onto microscope slides and imaged at 20x magnification using the ZEISS Slide Scanner (Carl Zeiss AG, Oberkochen, Germany).

Whole slide images of each placenta were manually segmented into different layers (Fig. 3(a)–(c)). Histological data collection and whole-slide histology image analyses were conducted independently by different researchers to minimize bias. To analyze the effect of the RUPP treatment on the relative size of each layer, we computed the area ratio of each segmented region to the whole placenta. We then compared the area ratios between RUPP and control placentas using an independent two-sample t-test to determine statistical significance.

Figure 3:

Histological analysis of the rat placenta. H&E image of a rat placenta with segmentations of the (a) junctional zone in gold, (b), labyrinth zone in green, and (c) chorionic plate in purple. Examples of 200 μm x 200 μm insets from the (d) junctional zone, (e) labyrinth zone, and (f) chorionic plate.

To connect the QUS results with biological features in the placenta, we measured the cell nuclear diameter in each anatomical layer of the placenta and compared with the regional ESD values. For each layer in the segmented whole slide images, five non-overlapping ROIs (200 μm x 200 μm) were randomly selected using a random number generator, further reducing potential bias in region selection (Fig. 3(d)–(f)). For each region of interest, the diameter of each cell nucleus was manually measured along the short axis and recorded for further analysis. A two-way analysis of variance (ANOVA) was used to calculate the effect of the placental layer and surgical group on the nucleus diameter. Once the variables with significant effects on variance were identified, an independent two-sample t-test was used to calculate whether there was a statistically significant difference.

Correlation between QUS and Histology

Direct correlation of in vivo image data with ex vivo histology data for each placenta was unfeasible, so in vivo and ex vivo datasets were analyzed independently. It is difficult to track multiparous pregnancies due to fetal, intestinal, and respiratory motion which regularly reorients the uterine horn. The only definitive method to track individual placentas would be a physical tag, which is impractical and unspecific for labeling up to six placentas per animal.

To correlate the QUS results with our histology data, we plotted the average cell nucleus diameter against the average ESD in each placental layer for each animal. For each animal, we took the average cell nucleus diameter in each placental layer from all the ex vivo histology placentas belonging to that animal and computed an animal-level average across all the placental-level averages. A similar average of averages was then calculated using the in vivo QUS placentas to obtain an animal-level ESD average for each placental layer. Then, each placental layer for each animal was plotted to visualize the relationship between ESD and cell nucleus diameter across placental layers. Finally, we used a linear regression to quantify the strength of the correlation between ESD and cell nucleus diameter across placental layers.

Sample Size Determination

We performed a statistical power analysis using G*Power software (35, 36), based on detecting differences between two independent means with a significance level (α) of 0.05 and power (1 - β) of 0.80. Using previously reported the means and standard deviations for the junctional zone area ratio in control and RUPP placentas from the results of our ex vivo histological analysis, we calculated that at least 12 placentas per group would be required. Considering that rats are multiparous animals and bear multiple placentas, we planned our histological analysis to include three placentas per animal, requiring at least four animals per experimental group. Therefore, we included four animals in the control group and five animals in the RUPP group. We included an additional animal in the RUPP group due to the known risk of fetal resorption resulting from the stress of the RUPP surgery, potentially reducing the final placenta yield. Furthermore, to account for potential exclusions from motion artifacts, we collected in vivo QUS data from more than three placentas per animal. The resulting sample sizes were control n = 18 and RUPP n = 20 for the in vivo QUS analyses and control n = 12 and RUPP n = 15 for the ex vivo histological analyses.

Results

Whole Placenta QUS Analysis

QUS analysis of the whole placenta demonstrated significant differences in average ESD values between RUPP and control placentas. Fig. 4(a) shows that average ESD across the placenta was 1.0 μm smaller (p = 0.040) in RUPP placentas (24.2 μm) compared to control placentas (25.2 μm). See Fig. S1 in the Supplementary Materials for whole placenta QUS analyses involving other QUS parameters.

Figure 4:

QUS provides image contrast that helps to differentiate between RUPP and control placentas. (a) Violin plot comparing the average ESD in control placentas (n = 18) and RUPP placentas (n = 20), *: p < 0.05. B-mode images of (b) control and (c) RUPP placentas show little contrast differentiating one another, while ESD parameter maps of (d) control and (e) RUPP placentas provide more contrasting information. Violin plots comparing the fraction of the placenta occupied by the (f) junctional zone, (g) labyrinth zone, and (h) chorionic plate between RUPP placentas (n = 15) and control placentas (n = 12), *: p < 0.05.

Regional Placenta QUS Analysis

Regional placenta QUS analysis found significant differences in ESD measurements within each anatomical layer of the placenta. The average ESD was significantly different between all three anatomical layers (Fig. 5(b)), with p = 5.6e-4 between junctional and labyrinth zone, p = 8.0e-3 between labyrinth zone and chorionic plate, and p = 4.5e-8 between junctional zone and chorionic plate. See Fig. S2 in the Supplementary Materials for regional placenta QUS analyses involving other QUS parameters.

Figure 5:

Cell nucleus diameter correlates with ESD. (a) Sampling curves for junctional zone, labyrinth zone, and chorionic plate in ESD parameter maps along with (b) violin plot comparing their values within each placenta (n = 38). (c) H&E image depicting the junctional zone (gold), labyrinth zone (green), and chorionic plate (purple) along with (d) violin plot comparing cell nucleus diameter measurements in each region for all placentas (n = 27). (e) Scatter plot correlating the cell nucleus and ESD values for each mother. ****: p < 1.0e-4, ***: p < 1.0e-3, **: p < 1.0e-2.

Two-way ANOVA showed that preeclampsia (RUPP vs control) had a significant overall effect on ESD across placental layers (p = 0.0028); when examining each layer individually, only the chorionic plate demonstrated a statistically significant difference in ESD between groups. The chorionic plate in RUPP placentas had a significantly smaller ESD compared to the control (p = 0.033), while the junctional and labyrinth zones showed no difference (p = 0.24 and p = 0.076, respectively), see Fig. S3(a) in the Supplementary Materials.

Ex vivo Histological Analysis

Histological analysis found significant differences between the average nucleus diameter in each placental segment (Fig. 5(d)), with p = 3.3e-21 between junctional and labyrinth zones, p = 3.1e-13 between labyrinth zone and chorionic plate, and p = 9.5e-27 between junctional zone and chorionic plate.

Two-way ANOVA showed that preeclampsia did not have a significant effect on cell nucleus size (p = 0.93) within each placental layer.

Histological analysis demonstrated that RUPP placentas had significantly smaller junctional zones compared to control placentas (Fig. 4(f), p = 0.013). This agrees with previous results in the literature, confirming the effects of the RUPP model. We also found that RUPP placentas had significantly larger chorionic plates compared to the control (Fig. 4(h), p = 0.031). The size of the labyrinth zone was not significantly different between surgical groups (Fig. 4(g), p = 0.070).

Correlation between QUS and Histology

Linear regression showed a good correlation (R² = 0.58, p = 3.8e-6) between the average cell nucleus diameter and average ESD (Fig. 5(e)).

Discussion

Our objective was to use QUS to detect the microstructural changes in the placenta occurring in preeclampsia, a life-threatening pregnancy disorder. Using the established RUPP rat model of preeclampsia, we correlated changes in QUS parameter maps of the placenta with changes in histology. In particular, maps of QUS parameter ESD showed significant differences between RUPP and control placentas, as well as significant differences between the anatomical layers of the placenta.

We demonstrated that RUPP placentas had significantly smaller average ESD compared to control placentas. The smaller ESD value in RUPP placentas may originate from their smaller junctional zones and larger chorionic plates. The junctional zone exhibited the largest cell nuclei compared to the other regions of the placenta while the chorionic plate featured the smallest cell nuclei. With a smaller junctional zone and a larger chorionic plate, RUPP placentas contained fewer cells with larger nuclei and more cells with smaller nuclei, which may have contributed to the decrease in overall ESD. Significant differences between control and RUPP placentas were detected in both the junctional zone and chorionic plate, although no significant differences were observed in the labyrinth zone. The labyrinth zone serves as the principal site of maternal-fetal nutrient and gas exchange (23), and thus its structural stability under acute ischemic conditions—as induced by the RUPP model—may indicate a degree of adaptive resilience specific to this region. Moreover, it is likely that the relatively short duration of placental ischemia in our animal model (four days before imaging) may not have provided sufficient time for the labyrinth zone to undergo measurable structural alterations detectable by current QUS methods. Our findings emphasize the sensitivity of QUS imaging in capturing rapid structural changes in specific placental regions while highlighting the labyrinth zone’s potential temporal or physiological differences in response to ischemia. Future research incorporating longer ischemic durations, combined imaging techniques, or molecular analyses will help further elucidate the labyrinth-specific responses and strengthen the translational relevance of these findings for clinical diagnosis and management of human preeclampsia.

We also found that cell nucleus diameter correlated well with ESD measurements throughout the placenta. Although we found a good correlation between ESD and cell nucleus, ESD values were approximately 3–6 times larger than the histology-based nucleus diameter measurements. Acoustic scattering in biological tissues is a complex phenomenon, so it is challenging to surmise exactly which anatomical structures are influencing the QUS parameters. Because acoustic scattering occurs at interfaces of materials with acoustic property mismatches, both solid structures and the fluid spaces between them will be sources of scatter. While QUS parameter values may not necessarily be influenced by a single structure such as the cell nucleus, we chose to correlate the QUS parameters with measurements of cell diameter because it was a discrete measurement which clearly varied between the different cell types present in the different layers of the placenta. In addition to cell nuclei, sources of scattering in placenta could include the blood in the vascular spaces, cytoplasm in the various cell types, and extracellular matrix proteins. The additional sources of scattering may be larger than the cell nuclei, which would influence the ESD estimates. ESD provides a general estimate of the size of scatterers within tissues, but larger scatterers tend to bias measurements and result in increased ESD values in broad distributions of different scatterer sizes (37). Another key reason for discrepancies between ESD and cell nucleus size is the chosen imaging frequency. This study employed a transducer with a 21 MHz center frequency, which is not as sensitive to structures as small as cell nuclei compared to the larger tissue components, such as cells, cell clusters, vascular spaces, etc.

Biological features such as cell size and vascular space in the placenta were more difficult to quantify using our methods. For instance, the syncytiotrophoblast is a major component of the labyrinth zone, but this mass of cell matter extends throughout the entire 3D structure of the labyrinth. The 3D structure of the syncytiotrophoblast makes it difficult to discretely measure in a 2D histological cross-section of the tissue. Similarly, vascular spaces in the labyrinth are also difficult to discretely measure with 2D histology images due to their extensive, complex 3D structure. Furthermore, increased presence of red blood cells in vivo likely affects the size of the vascular spaces, meaning that measuring vascular spaces with histology may not provide an accurate depiction of the in vivo condition. This study did not focus on quantifying any aspect of red blood cells, as these cells do not contain cell nuclei; however, red blood cells likely also contribute to scattering. Measuring the effects of red blood cells on scattering using ex vivo tissue is challenging, as these cells are present in much larger amounts in vivo. Alternative 3D tissue imaging techniques will likely be necessary to accurately quantify the contributions of cells, blood, and vascular spaces to scattering phenomena.

We chose to focus on the junctional zone, labyrinth zone, and chorionic plate for our analyses of the placenta. The placenta also includes other anatomical features such as the decidua, yolk sac, and metrial glands; however, we found that these features were not as reliable for image analysis. The metrial gland was rarely present in our dissected placenta tissue samples, making it difficult to draw any biological conclusions about this anatomical feature. The decidua and yolk sac were often present in dissected placental samples; however, these layers are so thin compared to the other layers of the placenta and so small relative to the QUS sampling ROI that their influence on the QUS parametric maps was likely negligible.

We found that the chorionic plate in RUPP placentas had smaller ESD compared to control but found no difference in cell nucleus diameter in the same region. The chorionic plate is the smallest of the three regions that we analyzed, meaning that it took up the smallest percentage of the QUS sampling ROI. Other structures present in the sampling ROI may have influenced our ESD estimates in the chorionic plate. Histological analyses found that RUPP placentas had larger chorionic plates compared to the control, meaning that control placenta chorionic plate ESD values were more influenced by surrounding structures. The size of the QUS resolution cell relative to the size of the chorionic plate may have introduced bias. While cell nucleus diameter in the chorionic plate may not change in the RUPP model, other anatomical features in the chorionic plate may be changing and therefore influencing the ESD measurements in that region. More in-depth analysis of the chorionic plate is required to draw more definitive conclusions. Since B-mode images did not provide sufficient contrast between the different layers of the placenta, we had to use anatomical landmarks present in the B-mode images to approximate the locations of the junctional zone, labyrinth zone, and chorionic plate when sampling our QUS parameter maps. The approximated sampling locations were placed as close as possible to the reference landmarks to ensure the accuracy of the sampling from each target layer; however, the QUS resolution cell size used to sample the radiofrequency data was almost always wider than the cross-section of the layer we were interested in sampling. While the large QUS resolution cell size increases the accuracy of scatterer size estimates, it decreases the spatial resolution of the QUS parameter maps. For instance, samples chosen to represent the junctional zone likely included information from the junctional zone as well as surrounding tissues such as the decidua, uterus, and even labyrinth zone. Similarly, the chorionic plate estimates may have included information from the amniotic fluid, umbilical cord, and labyrinth zone as well. While the junctional zone and chorionic plate had specific anatomical landmarks to help locate them, the labyrinth zone had no such landmarks; however, the histology results demonstrated that the labyrinth zone consistently occupied more than half of the placenta’s total cross-sectional area, providing a large margin in the center of the placenta for our B-mode approximations.

We acknowledge that translating our findings from a preclinical rat model to clinical human applications involves certain challenges, including differences in placental anatomy and microstructure between species. Additionally, the use of lower-frequency clinical ultrasound transducers required for imaging deeper human placentas may impact imaging resolution and sensitivity to smaller scatterers, potentially influencing the diagnostic capability of QUS methods. Nevertheless, prior studies have successfully demonstrated effective clinical translation of QUS techniques from high-frequency preclinical imaging to lower-frequency clinical applications in various other tissues (15, 16, 21, 22). While no animal model is a perfect representation of the human condition, this animal study establishes a method for placental imaging with QUS that could easily translate to a clinical application. Although there are anatomical differences between human and rat placentas, both organism’s placentas exhibit significant microstructural differences with the reduction of placental perfusion. It likely that the human placenta will also exhibit changes in QUS parameters, but these changes may require a different analysis approach to detect differences caused by preeclampsia. Although careful optimization and validation are needed, we remain confident in the clinical potential of QUS for evaluating human placental microstructure and aiding early diagnosis and management of preeclampsia.

Integration of QUS methods into current preeclampsia screening methods may allow for earlier detection of the disease. QUS can be easily integrated into current clinical imaging workflows with minimal change to the clinician and technician workflow. Many clinical US systems allow the collection of RF data. QUS parameters would be processed and then used to determine whether the patient is developing preeclampsia using a very intuitive easy-to-interpret display. QUS could also be used to monitor treatment or assess the severity of the condition. For instance, our group is currently developing therapeutics to treat the symptoms of preeclampsia, and earlier diagnosis with QUS would allow for earlier treatment of the condition. Furthermore, QUS could also provide a method for non-invasively monitoring treatment progress.

Conclusion

Preeclampsia is a life-threatening pregnancy condition that is poorly understood, but QUS methods show great promise as an imaging tool for detecting changes in the placenta caused by the disease. This study validated QUS as a tool for detecting microstructural differences in rat placentas. QUS can distinguish between normal and diseased placentas, as well as detect differences between the anatomical regions of the placenta. Future studies will initially involve validating our QUS methods using ex vivo human placenta samples collected postpartum, comparing normal and preeclamptic placentas to establish clinically relevant baseline parameters. Subsequent studies will translate these optimized methods into clinical research, involving prospective longitudinal QUS imaging of placental development in pregnant individuals at risk for preeclampsia. This approach will enable non-invasive monitoring of placental microstructural changes in real-time, potentially facilitating earlier and more accurate identification of at-risk pregnancies. Ultimately, this research aims to advance clinical practice by providing physicians with reliable imaging biomarkers for preeclampsia, leading to improved maternal and fetal health outcomes through timely diagnosis and intervention.

Data Availability

The US RF echo data used in this study is confidential and cannot be made freely available.