Skip to main content
NCBI home page
Journal of Anatomy logoLink to Journal of Anatomy
. 2024 Sep 10;246(1):134–147. doi: 10.1111/joa.14131

Three‐dimensional virtual histology of the rat uterus musculature using micro‐computed tomography

Mathias W Roesler 1, Amy S Garrett 1, Mark L Trew 1, Dane Gerneke 1, Satya Amirapu 2, Leo K Cheng 1, Alys R Clark 1,✉
PMCID: PMC11684386  PMID: 39253979

Abstract

Contractions of the uterus play an important role in menstruation and fertility, and contractile dysfunction can lead to chronic diseases such as endometriosis. However, the structure and function of the uterus are difficult to interrogate in humans, and thus animal studies are often employed to understand its function. In rats, anatomical studies of the uterus have typically been based on histological assessment, have been limited to small segments of the uterine structure, and have been time‐consuming to reconstruct at the organ scale. This study used micro‐computed tomography imaging to visualise the muscle structures in the entire non‐pregnant rat uterus and assess its use for 3D virtual histology. An assessment of the rodent uterus is presented to (i) quantify muscle thickness variations along the horns, (ii) identify predominant fibre orientations of the muscles and (iii) demonstrate how the anatomy of the uterus can be mapped to 3D volumetric meshes via virtual histology. Micro‐computed tomography measurements were validated against measurements from histological sections. The average thickness of the myometrium was found to be 0.33 ± 0.11 mm and 0.31 ± 0.09 mm in the left and right horns, respectively. The micro‐computed tomography and histology thickness calculations were found to correlate strongly at different locations in the uterus: at the cervix, r = 0.87, and along the horn from the cervical end to the ovarian end, respectively, r = 0.77, r = 0.89 and r = 0.54, with p < 0.001 in every location. This study shows that micro‐computed tomography can be used to quantify the musculature in the whole non‐pregnant uterus and can be used for 3D virtual histology.

Keywords: micro‐computed tomography, muscle fibres, myometrium, uterus, virtual histology

Micro‐CT imaging of the uterus, stained with phosphotungstic acid, allows high‐resolution imaging of the organ. Tissue regions (myometrium and endometrium) are distinguishable in the 3D reconstruction of the organ, and muscle thickness and fibre orientation can be quantified from micro‐CT imaging. Observations were confirmed using histology, paving the way for future studies to quantify variation in organ anatomy in health and disease.

graphic file with name JOA-246-134-g002.jpg

1. INTRODUCTION

The uterus is crucial for mammalian reproduction. Cyclical changes in levels of hormones, contractility and excitability occur during the oestrus cycle, which occurs in most mammals (Wray & Noble, 2008). During this cycle, the uterus prepares for pregnancy by increasing the vascularisation of the endometrial lining (Clark & Kruger, 2017). If an oocyte is fertilised, the uterus continues to change and adapt to host the foetus during the gestation period. Non‐pregnant uterine contractile activity helps shed the menstrual lining during menses (in menstruating mammals) and facilitates sperm transport and implantation (Kunz & Leyendecker, 2002), and so plays a significant role in fertility. Perturbations in uterine contractility through the menstrual cycle can cause chronic conditions. For instance, endometriosis is characterised by endometrial‐like cells outside the uterus and affects up to 10% of women (Becker et al., 2022). The implantation and proliferation of these cells, which have been linked to retrograde menstruation, create endometriotic lesions associated with pelvic pain, dysmenorrhea (painful menstruation) and infertility (Tosti et al., 2015).

For ethical reasons and due to the lack of accessibility of the human uterus, studies of uterine function primarily use laboratory animals. Rodents, which are commonly used in studying uterine function (Garrett, Means, et al., 2022; Garrett, Roesler, et al., 2022), have bicornuate uteri composed of three layers, from the lumen outwards: the endometrium, the myometrium and the perimetrium (Malik et al., 2021). The myometrium comprises smooth muscle cells (SMCs), organised in fibres with clear orientations in rodents such as rats and mice. The outer layer of the uterine horn is composed of longitudinal fibres, whereas the inner layer is made of circumferentially arranged fibres (Malik et al., 2021). Bridge‐like structures have also been identified in pregnant rats and non‐pregnant mice, linking the longitudinal and circumferential muscle layers together (Kagami et al., 2020; Lutton et al., 2018).

Contraction of uterine smooth muscle is caused by an influx of calcium (Ca2+) into the SMCs (Garfield & Maner, 2007). Spontaneous electrical activity can be measured, in the form of action potentials, at the tissue level by placing electrodes on the surface of the organ (Garrett, Means, et al., 2022; Garrett, Roesler, et al., 2022). The electrical properties of the muscle layers vary. Action potentials travel faster in the longitudinal layer and have a greater amplitude than those in the circumferential layer (Osa & Fujino, 1978; Rabotti & Mischi, 2015). Computational models of the human uterus indicate that fibre orientations and the location of imposed pacemaker sites impact the propagation of electrical events more than geometry (Zhang et al., 2016). Thus, more information about muscle structures and electrophysiology is needed to improve our understanding of uterine contractility.

Histology offers a two‐dimensional view of the micro‐structure and muscle layers in the uterus (Bergman, 1968; Jain et al., 2000). However, this method only provides local information and cannot capture variations across the entire horn. Lutton et al. (2017) constructed a three‐dimensional view of the smooth muscles of an entire pregnant rat uterus and a block (70 × 30 × 3.5 mm) of a non‐pregnant human uterus using over 200 serial histological slides. The final resolution of their reconstruction was approximately 50 μm/voxel. Computational methods were specifically developed to visualise and quantify fibre orientations and micro‐structures in the uterus. In their study, Lutton et al. (2017) were able to detect distinct circular and longitudinal muscle layers in the uterus, at high resolution. However, computational reconstruction of histological segments requires careful alignment of images and can be time‐consuming.

Recent studies on 3D virtual histology have shown that micro‐computed tomography (μCT) imaging of PhosphoTungstic Acid (PTA)‐stained hearts and kidneys provides similar results to those obtained with classic histology methods, with the ability to create a three‐dimensional reconstruction of the imaged organ (Dullin et al., 2017; Missbach‐Guentner et al., 2018). This method, which has yet to be applied to the uterus, could provide information on the muscle layers in the entire uterine horn in three dimensions. A whole organ view of the muscle layers can help develop anatomically realistic models of the uterus, which are an essential tool in understanding the electrophysiological and mechanical activity of the organ (Garrett, Means, et al., 2022; Garrett, Roesler, et al., 2022).

In this study, μCT was used to characterise the muscular structures of the entire non‐pregnant rat uterus, from the cervix to the ovaries. The PTA stain enhanced contrast in the images and facilitated the distinction between myometrium and endometrium. Histology was performed on one of the horns to validate using PTA and μCT imaging for 3D virtual histology of the rat uterus. Additionally, the μCT images were used to extract the fibre orientations of the muscle layers and create an anatomically realistic volumetric mesh of the organ. Studying the muscular structures and their organisation in more detail will help improve our understanding of the contractile activity occurring outside of pregnancy.

2. MATERIALS AND METHODS

2.1. Sample preparation

Ethical approval was obtained from the University of Auckland Animal Ethics Committee, and all procedures followed the ethics guidelines. One healthy female Sprague Dawley rat (11 weeks old, 320 g) was euthanised via cervical dislocation while under anaesthesia, and its uterus was excised immediately. The rat was determined to be in the oestrus phase of its cycle by visual assessment (Ajayi & Akhigbe, 2020). The left horn of the uterus was marked with a suture, as seen in Figure 1a. The organ was pinned down on a styrofoam block in its approximate in vivo arrangement by placing pins in the fatty tissue attached to the uterus before being submerged in a 2% formalin solution for 24 h. The sample was then placed in phosphate‐buffered saline (PBS) for 18 h to flush out excess formalin. The organ was dehydrated in three different ethanol baths of increasing concentration. Each bath lasted 2, 4 and 16 h, and the concentrations were 30%, 50% and 70% ethanol, respectively. After the dehydration process, the fatty tissue was removed with a scalpel. Finally, the sample was stained using a 2% phosphotungstic acid (PTA) solution for 12 days. During the entire preparation process, the sample was consistently submerged in solution and placed on a Stuart SSM4 mini see‐saw rocker (Cole‐Parmer, Vernon Hills, IL) at a frequency of 0.25 Hz to help the diffusion of the solutions in the entire sample.

FIGURE 1.

FIGURE 1

Open in a new tab

Workflow of the methods used to process the uterus. (a) Freshly excised uterus in the oestrus stage of its cycle from a female Sprague–Dawley rat. The blue suture was used to identify the left horn. (b) After being fixed and stained, the entire uterus was imaged using micro‐computed tomography (μCT), and a three‐dimensional reconstruction was generated. This dataset was then processed to segment the myometrium from the endometrium. These segmentation masks were used to estimate muscle thickness, generate a volumetric mesh of the myometrium and analyse fibre orientations. (c) The uterus after being stained and imaged once. Both horns were then sectioned into smaller segments. The left horn was used for high‐resolution μCT imaging and the right horn for histology. (d) View of one slice from a small segment of the left horn imaged with μCT. (e) Histology slide stained with Hematoxylin and Eosin from a section of the right horn.

2.2. μCT imaging

The uterus was imaged twice with a Brucker 1272 Skyscan instrument (Bruker, Kontich, Belgium). The first scan was a whole organ scan at a resolution of 10.9 μm/voxel to capture the entire uterus and provide a general view of the muscle structures along both horns. The entire sample was placed vertically in a tube and held in place with two styrofoam pieces to prevent the sample from moving during the imaging.

The second scan was performed on smaller segments from the left horn and the cervix at a higher resolution of 1.5 μm/voxel. The horn was sectioned into three pieces: an ovarian segment close to the ovaries, a central segment and a cervical segment near (but not within) the cervix. Each segment was ∼2 mm long and ∼2 mm wide. All the segments were stacked in a plastic tube and separated by small styrofoam pieces.

Scotch® Aluminium Foil Tape 3311, which attenuates low‐energy x‐rays, was placed in front of the source for both scans to act as a filter. A flatfield correction was performed prior to each scan. The list of scan parameters that were used for each imaging is presented in Table 1. The images were reconstructed using InstaRecon CBR Server Premium 15K (InstaRecon Inc., Champaign, IL) with NRecon 2.2.1 (Bruker, Kontich, Belgium) and visualised in 3D using CTVox V 3.3 (Bruker, Kontich, Belgium). All datasets are available online (see Supporting information S1).

TABLE 1.

List of scan parameters for the μCT imaging of the rat uterus.

Full organ Small segments
Energy (kV) 92 98
Current (μA) 108 102
Exposure time (ms) 1800 3400
Rotation step (over 360°) 0.35° 0.25°
Voxel size (μm) 10.9 1.5
Number of fields 2 1 per segment
Number of pixel (per field) 2452 × 1640 4904 × 3280
Scan time (hours per field) approximately 3 approximately 3
Open in a new tab

2.3. Histology

The right uterine horn and part of the cervix were subsequently used for histology. After being prepared and imaged using μCT, the uterus was divided into sections for histological staining and imaged at four different locations: (i) the ovarian end of the horn, (ii) the centre of the horn, (iii) cervical end of the horn and (iv) the cervix. The uterine segments were placed in cassettes before being embedded in paraffin. Slices were cut from each block with a 5 μm thickness and subsequently stained with Hematoxylin and Eosin (HE). The slices were imaged with a Digitech QC3199 microscope with a resolution of 3.2 μm/pixel. The complete dataset is available online (see Supporting information S1).

2.4. Data processing

2.4.1. Muscle segmentation

The μCT images were processed to segment the myometrium from the endometrium. The transverse slices of the entire uterus dataset were first downsampled by a factor of 4 for computational efficiency. The new dataset contained 455 slices of 1124 × 548 pixels, which maintained isometric voxels with a resolution of 43.77 μm/voxel, comparable to previous histological studies. The horns were processed separately, and the analysis was performed on 455 slices for the left horn and 305 for the right horn. The ovaries from both horns were removed from further processing. The images were segmented using the MATLAB 2022b (The Mathworks, Inc.) k‐means algorithm from the Statistics and Machine Learning Toolbox, with three clusters and a threshold value of 1 × 10−4 to separate the background, endometrium and myometrium. From the myometrial cluster, segmentation masks were created, which were subsequently processed manually using ITK‐SNAP (version 4.0.0, http://www.itksnap.org) to refine the segmentation and correct small regions mislabelled as myometrium or parts of the myometrium that were not labelled. Each slice was inspected, and small modifications were made to most slices; however, the semi‐automated method provided a more streamlined annotation of images than manual segmentation of each slice. An example of one slice and its mask are shown in Figure 2a,b, respectively. The segmentation masks were then used to determine the thickness of the muscle wall from the cervix to the ovaries, create anatomically realistic meshes and to extract fibre orientations in the myometrium. The histology slides were segmented manually using ITK‐SNAP and are available online (see Supporting information S1). Three observers conducted manual segmentations of μCT and histology slices to confirm reliability in manual and semi‐automated segmentations. Similarity between segmentations was analysed using the segmentationmetrics library in Python (https://pypi.org/project/segmentationmetrics/).

FIGURE 2.

FIGURE 2

Open in a new tab

The steps of the algorithm to estimate the muscle layer thickness. (a) Slice from the micro‐computed tomography (μCT) dataset showing both uterine horns. The muscle layers are light grey and surrounded by the endometrium, which appears darker. (b) Segmentation mask of the muscle layers from the slice shown in (a). (c) Segmentation mask after being rotated to create a transverse slice of the right horn. The centreline of the right horn was aligned with the z‐axis of the stack of images from the μCT dataset. (d) Reoriented segmentation mask shown in (c), with the centre point of the right horn shown in red and the projection point pairs in eight directions shown in blue. The anti‐mesometrial border is identified with the purple point. Each pair had one point on the outer edge and one point on the inner edge of the muscle layers.

A similar process was applied to the higher resolution images from the smaller ovarian segment. They were downsampled by a factor of 4, thus creating a new dataset with 325 slices of 753 × 518 pixels and a resolution of 6.0 μm/voxel. The μCT images were segmented manually using ITK‐SNAP and used for fibre analysis. Both downsampled datasets and the segmentation masks are available online (see Supporting information S1).

2.4.2. Thickness analysis

Both horns were processed independently to estimate the thickness of the myometrium. Each segmentation mask of the μCT dataset was rotated to create transverse slices of the horns. A centreline was estimated for the uterine tissue. First, the centroid of the region of interest comprising endometrial tissue/cavity was calculated from a segmentation generated to describe the endometrium‐myometrium border. Then, the centreline was approximated with the vectors between centroids. Finally, the segmentation masks were rotated to align the centre vector with the z‐axis of the μCT image stack, creating transverse slices where the centreline of the horn was normal to the image plane. Figure 2c shows the muscle layer mask after rotating the slice to align the right horn.

The thickness of the myometrium was estimated by calculating the distance between points on the inner and outer edges of the myometrium. The cervix was divided in half to allow the left and right canals to be processed separately. Pairs of points were found along lines crossing the centroid of the mask at different angles at the intersection with the inner and outer borders. The anti‐mesometrial border was used as a landmark and defined at the angle θ = 0 rad. The total circumference of the uterine horns was covered with 256 pairs of points. Figure 2d shows an example with eight pairs of points in blue and the centroid in red. The Euclidean distance between the outer and inner points was computed for each of the 256 pairs to determine the thickness of the muscle layers. The values were averaged to obtain a single thickness value for each slice.

The same method was used for the segmentation masks of the histology slides and the high‐resolution ovarian segment without performing any rotation. To compare the data from the histology and the μCT, the Pearson correlation coefficients were calculated between the thickness values of the histology slices and four μCT slices at similar locations in the horn using the SciPy stats library (https://docs.scipy.org/doc/scipy/reference/stats.html).

2.4.3. Anatomically realistic mesh generation

Surface meshes of the myometrium of the whole uterus and the ovarian segment with triangular faces were generated from the segmentation masks using the surface mesh generation algorithm of ITK‐SNAP. The surface meshes were registered with Cloud Compare (version 2.12.4, https://www.cloudcompare.org) to compare the data at different resolutions (see Supporting information S2). Only the whole organ surface mesh was used to generate a volumetric mesh. Isolated triangles that were not part of the mesh were removed using Meshmixer (version 3.5.474, Autodesk, Inc.), and areas that contained holes were filled. Finally, Netgen (version 6.2.2304, https://ngsolve.org) was used to generate a volumetric mesh with tetrahedral elements. Additionally, the thickness information was mapped to the volumetric mesh. The centreline was used to slice the mesh perpendicularly to the horns, and the elements that intersected the plane were annotated with the thickness value associated with that slice number. The different surface and volumetric meshes are available online (see Supporting information S2).

2.4.4. Fibre analysis

Muscle layer fibre orientations were determined with 3D structure tensor analysis of the μCT dataset segmentations of the entire uterus and of the ovarian segment. The segmentation masks were first upsampled by a factor of 4 to match the resolution from the original datasets (10.9 μm/voxel for the whole organ and 1.5 μm/voxel for the smaller segment). Muscles were extracted from the original μCT images using the upsampled segmentation masks so that only image information of interest was retained for structure tensor analysis. To minimise tissue‐background boundary effects, the masked images were eroded along boundaries using a spherical structuring element (r = 9 voxels), and iterative diffusion filters were used to smooth the tissue signal into the image background while retaining the original detail in the masked region (Gilbert et al., 2012). Briefly, a 3D diffusion filter was constructed and iteratively applied to the image. Using the boundary diffused image, optimal gradient filters (Farid & Simoncelli, 2004) were efficiently applied using circular convolution and fast‐Fourier transforms. Specifically, the gradient at all voxels in image coordinates i,j,k was found by convolving a vector of gradient weights, χ, with the image reshaped into a vector, f. This is equivalent to the component‐wise product of 1D fast‐Fourier transforms, F (Dongarra et al., 2000):

gi=χi*f=F−1Fχi∙Ff

where F−1 is the inverse fast‐Fourier transform. This is an efficient computation with the product found in Onlogn operations. It has been implemented in MATLAB 2022b (The Mathworks, Inc.), and all code is available on GitHub (see Data S1). At each voxel, the six unique components of the structure tensor are the outer product of the gradient vectors in each image coordinate (Rutherford et al., 2012):

ST=∇I∇IT=gi2gigjgigkgjgigj2gjgkgkgigkgjgk2

where I is the μCT intensity image and gi, gj and gk are the intensity gradients in image coordinates i, j and k, respectively.

Structure tensor components were spatially smoothed for robust non‐local fibre orientation predictions (Brox et al., 2006; Jähne, 2005). Here, multi‐level smoothing was used to isolate dominant fibre features across scales of interest (Rutherford et al., 2012). A monophasic five‐point 1D binomial filter (Jähne, 2005) was iteratively applied to efficiently construct a power‐of‐2 hierarchy of structure tensor components (Rutherford et al., 2012). The fourth smoothing scale level was used to construct structure tensors and perform eigenanalysis, providing data resolutions of 174.4 and 24 μm for the whole organ and ovarian segment, respectively. The eigenvector corresponding to the smallest eigenvalue of the structure tensor captured the fibre orientation. Voxel‐wise eigen‐analysis was performed using parallel threads.

3. RESULTS

A 3D reconstruction of the uterus based on the μCT images is shown in Figure 3a. It was possible to distinguish between the myometrium and endometrium based on the greyscale intensity differences between the voxels representing each tissue in the images. The regions with higher intensity absorbed a larger volume of PTA during the staining process. PTA binds to fibrin and collagen (Quintarelli et al., 1971), making the myometrium, due to the smooth muscle fibres present, a denser region relative to the endometrium. Figure 3b depicts the colour‐coded intensity variations: the myometrium is the highly stained region appearing in yellow, whereas the endometrium absorbs less stain and therefore appears purple. The cervical region is composed of muscle and connective tissue rich in collagen (Harkness & Harkness, 1959), making it harder to distinguish the myometrium from the rest of the tissue in the cervix.

FIGURE 3.

FIGURE 3

Open in a new tab

Three‐dimensional reconstruction of the entire rat uterus from the micro‐computed tomography data. (a) Greyscale view of both horns showing the cervix and the ovaries. (b) A longitudinal slice of the entire uterus is colour‐coded to visualise the density differences. Regions with a higher density of PTA, such as the cervix, appeared yellow/red, whereas regions with less PTA were purple. The myometrium was visible along the horns in yellow/red and the endometrium in purple. The cervix absorbed larger quantities of PTA due to the greater presence of collagen amid the muscles.

Figure 4 shows the three longitudinal planes of the same section of the uterine horn with different depths in the reconstruction of the segment of tissue near the ovaries. The different layers were identified with white dashed lines: longitudinal muscle (LM), vascular layer (VL) and circumferential muscle (CM). In Figure 4a, only the longitudinal muscle fibres were visible and oriented from top to bottom. Figure 4b shows the transition between the longitudinal and circumferential muscle layers. Some structures can be observed within this transition zone without any distinct orientation. These could be bridging fibres between the two muscle layers; however, they could also be blood vessels located near the mesometrial border. Figure 4c depicts the circumferential muscles with fibres oriented in a perpendicular direction relative to the longitudinal ones.

FIGURE 4.

FIGURE 4

Open in a new tab

A view of the three distinct layers in the uterus from the high‐resolution micro‐computed tomography dataset of the segment located near the ovaries. The 3D volume was virtually sectioned longitudinally with the same plane at different depths. The different layers are identified with white dashed lines: longitudinal muscle (LM), vascular layer (VL) and circumferential muscle (CM). (a) The first section displayed the longitudinal fibres, in orange, which are mainly oriented parallel to the longitudinal plane. (b) The second section showed the transition between the longitudinal and circumferential muscles and contained blood vessels. Some fibrous structures were visible without any preferred orientation. The longitudinal muscles were still visible on the sides of the section. (c) The deepest section displayed the circumferential muscle layer at the centre, where the fibres were oriented perpendicularly to the longitudinal ones.

Inter‐observer error was quantified for each segmentation type. Three observers manually segmented a histology image, obtaining mean Dice scores of 0.94 (range 0.94–0.95), mean Jaccard scores of 0.89 (range 0.88–0.90) and Hausdorff distance of 1.5 pixels (range 0.2–2.2 pixels). The same three observers segmented a μCT image, obtaining mean Dice scores of 0.91 (range 0.89–0.92), mean Jaccard scores of 0.83 (range 0.80–0.86) and Hausdorff distance 1.8 pixels (range 1.4–2.0). The manual segmentations were also compared to the semi‐automated segmentation of μCT obtaining mean Dice scores of 0.85 (range 0.82–0.92), mean Jaccard scores of 0.74 (range 0.69–0.83) and Hausdorff distance of 1.7 pixels (range 1.0–2.0). Discrepancies in segmentation occurred primarily at the mesometrial border, where the longitudinal layer separates from the circumferential layer (resulting in an irregular shape, Figure 2), and there is subjectivity in the choices made in the extent of segmentations in this region.

Individual 2D slices from the full organ μCT stack were manually registered to the histological slices at different locations in the uterus. Figure 5 shows an overlay of the right horn from the μCT slices fitted to the histology slices in the three segments of the horn (A–C) and the cervix (D). Dashed lines were added to emphasise the separation between the myometrium and endometrium. Similarity was assessed between segmentations of μCT sliced perpendicular to the major axis of the horn and of histology sections registered to the μCT images manually. Dice scores for these comparisons were 0.85, 0.84, 0.83 and 0.84 for the ovarian, centre, cervical horn and cervix regions, respectively. Jaccard scores for these comparisons were 0.74, 0.73, 0.71 and 0.73 for ovarian, centre, cervical horn and cervix regions, respectively. Hausdorff distances were 3.2, 0.3, 0.4 and 0.5 pixels. As with manual segmentations, the primary sites of difference between the segmentations were in the mesometrial region, due to differences in the identification of separation between muscle layers. There were additional regions of the surface of the horn where histology did not exactly match to μCT (see Figure 5), which most likely relates to the manual registration process and difficulties in exact matching of the angle at which the horn is cut through to the histology, which may not be cut exactly perpendicular to the major axis of the horn. In the cervix, there are clear regions of discrepancy outside of the mesometrial region. Due to the cutting of the tissue to prepare for individual horns for imaging at high resolution, the region between the two cervical openings to the vagina could not be captured fully, and muscular tissue was not as clearly distinguishable in either μCT or histology as it was in the external layers of tissue. Rather than subjectively defining a region of muscle between the two horns, this region was excluded from subsequent analysis.

FIGURE 5.

FIGURE 5

Open in a new tab

Comparison between the 2D Hematoxylin and Eosin‐stained histological slices and the microcomputed tomography (μCT) at different locations along the horn. Individual 2D slices near the ovaries (a), in the centre (b), near the cervix (c) and in the cervix (d) from the μCT stack of the full organ were manually registered to the histology slices. Black dashed lines indicate the separation between the myometrium and the endometrium. The schematic shows the approximate location of the slices for each panel.

The μCT images were used to estimate the thickness of myometrium in the uterus, and the thickness values from the histological data were used as validation. Figure 6a shows how the average thickness of the muscle layers varied along each horn from the cervix to the ovarian end for both horns. In this sample, the myometrium was at its thickest in the cervix. The thickness decreased as the cervix transitioned into the separate horns, where it remained relatively constant along the length of the horns. Near the ovaries, the muscle thickness increases as the horns connect with the oviducts. Figure 6b,d show the variations in thickness of individual slices at different locations along the left and right horns, respectively. In the cervix, the region near the anti‐mesometrial border was the thickest; the thickness decreased near the shared border between the left and right canals. Thickness information derived from μCT of slices from the left and right horns showed that the thickness varied circumferentially, with a sharp increase in thickness at the anti‐mesometrial border near θ = 0 rad, compared to elsewhere. Approximately π rad from it, two smaller peaks were separated by a trough associated with the mesometrial border. The trough corresponded to the section with only circumferential muscle in the segmentations. Figure 6c depicts the variations in thickness of the right horn based on the segmentation from the histological slides. The plots show similar trends to those seen in (D). While there are differences in estimates of muscle thickness in regions where segmentations deviate, particularly in the region of the mesometrial border, in this region, the longitudinal muscle layer separates from the circumferential, and due to variation in the direction of this separation, the longitudinal layer is not considered in thickness calculations. However, overall, estimates of muscle thickness in both micro‐CT and histology are consistent, and variability in calculations in the mesometrial region does not significantly influence the overall average thicknesses calculated in each horn. The average thickness and standard deviation in the right horn were 0.31 ± 0.09 mm, and in the left horn 0.33 ± 0.11 mm. These results confirm that the myometrium in both horns was similar. The thickness in the ovarian segment was 0.31 ± 0.02 mm, which demonstrated a strong similarity between the low‐ and high‐resolution datasets. Using the segmentation from the histological slides, the average thickness was 0.31 ± 0.11 mm, consistent with μCT measurements. Additionally, Pearson correlation coefficients were computed between the thicknesses calculated from μCT data from the plots in Figure 6c and the histology data from Figure 6d. At each location, the data had a strong positive correlation with the correlation coefficients from the cervix to the ovaries, respectively: r = 0.87, r = 0.77, r = 0.89 and r = 0.54, all with p < 0.001.

FIGURE 6.

FIGURE 6

Open in a new tab

Variation of the muscle thickness along the uterine horns. The thickness of the myometrium was estimated from the reoriented segmentation masks for each horn. The cervix was divided into two sections along the centre to estimate the thickness on the left and right sides separately. (a) Plot of the average thickness along both horns. Both traces displayed similar trends, with the cervix being the thickest region and the horn having a relatively constant thickness. Near the ovaries, the thickness increased again as the horns closed. The error bars represented the standard deviation for 20 slices. (b) Plot of the variations in thickness for individual slices from the left horn. θ = 0 rad was set at the anti‐mesometrial border, which corresponded to the thicker region of the horn. Separated by approximately π rad, two smaller peaks and a trough corresponded to the mesometrial border, where there were only circumferential muscles in the horn. In the cervix, the shared border between the left and right canals was excluded from the thickness calculations. (c) Plot of the thickness variations using the histology data, which displayed similar trends to the micro‐computed tomography data. (d) Similar plot as in (b) for the right horn.

The downsampled μCT dataset of the organ was used to generate a volumetric mesh, which was annotated with the muscular thickness data. Figure 7a shows a frontal and sagittal view of the annotated mesh. In this sample, the cervix was the thickest region of the uterus, and the thickness of the muscles remained constant along the horns, except near the ovaries, where it increased. Figure 7b is a view of the mesh sliced longitudinally, which shows the thickness of the myometrium in the entire uterus.

FIGURE 7.

FIGURE 7

Open in a new tab

Meshes generated from the segmentation masks of the full uterus and coloured based on thickness. (a) Frontal and sagittal views of the organ volumetric mesh, which can be compared to direct 3D reconstructions from imaging shown in Figure 3. The tetrahedral elements were coloured based on the thickness of the slices. The cervix was divided into two sections along the centre to estimate the thickness on the left and right sides separately. The cervix was found to be the thickest region, and the thickness along both horns remained relatively constant. (b) View of the volumetric mesh after a longitudinal cut, which showed the thickness of the muscle walls along both the horns and the cervix.

Lastly, the segmentation masks of two μCT datasets were used to extract dominant fibre orientations using structure tensors. Figure 8a is a frontal view of the fibre orientations in the whole uterus dataset at the initial scan resolution of 10.9 μm/voxel. A zoomed‐in view of the segment near the ovaries shows the longitudinal and circumferential features. They were coloured to denote the angle relative to the centreline of each horn. Figure 8b shows a frontal view and a view rotated along the longitudinal axis of the fibres identified in the smaller segment located near the ovaries. They were coloured based on the angle with the XY plane. With the higher resolution images, the structure tensors identified more longitudinal and circumferential muscle fibres than with the whole uterus images; however, the whole uterus images demonstrate a scale‐invariant transition from flat (circular) to steep (longitudinal) fibre angles at the same location that is observed in histology images (Supporting information S3).

FIGURE 8.

FIGURE 8

Open in a new tab

Dominant orientations of fibre features identified using structure tensors on the full uterus and the smaller segment near the ovaries. (a) Frontal view of the fibres in the entire uterus with a zoom of the ovarian end of the horn. The structure tensors mainly identified the longitudinal structures along both horns; however, circumferential structures were also found closer to the interior of the muscle. They were coloured based on the angle relative to the centreline of each horn, making the outer longitudinal structures blue or green and the inner circumferential structures red or yellow. (b) Frontal view and a view rotated along the longitudinal axis of the fibres in a small section of the uterus. Using the higher resolution dataset of the segment near the ovaries, both longitudinal and circumferential orientations were clearly identified. They were coloured based on the angle relative to the XY plane, making the outer longitudinal fibres blue or green and the inner circumferential fibres red or yellow.

4. DISCUSSION

In this study, the capabilities of μCT imaging to quantify the structure of the rat uterus were demonstrated. Tissue was stained using PTA, known to bind to fibrin and collagen (Quintarelli et al., 1971) and previously used to show muscle fibres in the murine heart (Dullin et al., 2017). Histology was performed to confirm that the muscles were indeed stained and visible, and both Dice and Jaccard scores were greater than 0.7 in all comparisons made, suggesting good similarity between modalities and observers. The primary source of differences between histology and μCT arose in the mesometrial region, where there is separation between muscle layers, which impacts the Jaccard and Dice scores, but low Hausdorff distances (<3.2 pixels) support that good agreement exists overall. The use of μCT data provided additional three‐dimensional information, which enabled the study of the muscular architecture in the entire uterus instead of specific locations. The transverse slices from the reconstructed organ were used to quantify the thickness of the muscle layers in the cervix and along the horns. The thickness was relatively constant in the horns and greater in the cervical region. The average thickness in both horns was similar, with 0.31 ± 0.09 mm in the right horn and 0.33 ± 0.11 mm in the left. Data from the histological slices and the high‐resolution μCT images confirmed these results with an average thickness of 0.31 ± 0.11 mm and 0.31 ± 0.02 mm, respectively. The thickness information was subsequently mapped to the anatomically realistic volumetric mesh created from the μCT slices. Finally, the data confirmed previous results on fibre orientations in the uterus with an inner circular layer and an outer longitudinal layer of muscles.

Qualitative assessment of μCT image contrast of the PTA‐stained uterus showed a differentiation between myometrium and endometrium. When the whole uterus was scanned, the resolution was not sufficient to easily distinguish the different muscle layers. However, imaging smaller segments increased the resolution from 10.9 μm/voxel to 1.5 μm/voxel. At this higher resolution, a distinction between the longitudinal and circumferential muscle layers and fibre orientation within these layers became apparent. Between the longitudinal and circumferential layers, fibrous structures were identified with no distinct orientation. Kagami et al. (2020) and Lutton et al. (2018) reported the presence of bridging structures in the non‐gravid mouse and pregnant rat uterus, respectively. Kagami et al. (2020) identified a bridging muscle layer along the entire horn, whereas Lutton et al. (2018) observed a seemingly random distribution of bridges linking longitudinal and circumferential muscle layers. However, the space between the muscle layers has been found to contain blood vessels (Jain et al., 2000). Because blood vessel tissue also contains smooth muscle cells (Todd et al., 1983), the identified structure could be blood vessel walls that have absorbed PTA during the staining process.

Histological methods have previously been applied to the uterus to study muscle layers (Bergman, 1968), micro‐structure variations during the oestrus cycle (Westwood, 2008) and fibre orientations in three dimensions (Lutton et al., 2017). Here, histology was used as a comparative tool to explore the possibility of using slices from the μCT images for 3D virtual histology. One of the limitations of μCT was the resolution at which the samples were imaged. The resolution obtained with the smaller uterine segments was similar to what can be obtained with histology. The results presented here indicated that using small sections provided the same distinction between longitudinal and circumferential muscle layers as histology slides stained with HE. This suggests that μCT, in conjunction with PTA, can be used for 3D virtual histology to study muscular structures in the rat uterus. The clear distinction between the myometrium and endometrium in the transverse slices of the μCT datasets allowed for semi‐automated segmentation of the muscle layers. However, with sufficient labelled data, neural networks could be trained to perform automated segmentation (Moeskops et al., 2016). The masks were used to estimate the thickness of the muscle layers along the horns and cervix.

In the stomach, which similarly comprises longitudinal and circumferential muscle layers, the thickness of the muscle wall varies at different locations. Further variations may be linked to changes in volume (Di Natale et al., 2022). Similar changes and effects may occur in the uterus. The organ remodels itself during the oestrus cycle (Westwood, 2008) and pregnancy (Wray & Prendergast, 2019). These changes may be accompanied by variations in the thickness of the muscle layers linked to changes in the volume of the endometrium. The results presented here, from the uterus in the oestrus phase, showed slight variations but an overall consistent thickness throughout the entire organ. A comparison between the two horns showed similar trends and values in the thickness variations. This was confirmed by comparing the histological slices from the right horn and the μCT images of the smaller segments from the left horn, which showed few visual differences. Furthermore, the thickness data from the histology and the μCT were strongly correlated, which supports using μCT imaging for 3D virtual histology. From the μCT dataset and the histology study, the muscle thickness was found to vary between 0.1 and 0.7 mm in transverse slices of the horns, where the thickest region of the slices was the anti‐mesometrial border. A study on wound healing in the rat uterus indicated that, in the dioestrus phase, the thickness of the uterine wall, which included both endometrium and myometrium, was 0.69 ± 0.09 mm (Micili et al., 2013), which was of the same order of magnitude as the results presented here. The region incorporating vasculature between the two muscle layers, including at the anti‐mesometrial border, was included in the segmentation, meaning that our data reflects the total thickness, rather than individual layers.

The electrophysiological properties of the uterus are still under active investigation (Garrett, Means, et al., 2022). Bursts of electrical activity trigger contractions in the uterus (Wray & Noble, 2008). Different models have been developed to simulate these bursts at the cellular level (Testrow et al., 2018; Tong et al., 2014). However, they have yet to be included in larger, multi‐dimensional models. Multi‐scale modelling creates a virtual environment that can be used to investigate the propagation of electrical events and contractile responses. With the use of high‐resolution electrodes in pregnant (Lammers et al., 2015) and non‐pregnant (Garrett, Roesler, et al., 2022) rat uteri, more accurate data on the propagation of electrical events can be obtained in order to inform multi‐scale models. Previously, Lutton et al. (2017) proposed a method to reconstruct the uterus from histological slides. They applied it to entire pregnant rat uteri and segments from non‐pregnant human uteri. For the rat, the uterus was opened longitudinally along the anti‐mesometrial border. It was then embedded in paraffin and sliced to a thickness of 5 μm before being stained with HE. The method required the generation of hundreds of delicate histology slices, of which a large proportion were damaged and had to be discarded. The slices were subsequently processed to create a 3D reconstruction of the myometrium. However, the model did not fully capture the geometry of the organ since it was opened longitudinally along the anti‐mesometrial border to be imaged. An advantage of μCT is that the imaging process creates a three‐dimensional reconstruction of the entire sample in less than an hour, which can further be used to create anatomically realistic meshes of the musculature, as shown in this study. Extending this method to the different stages of the oestrus cycle or pregnancy could provide a model of structural changes in the uterus.

The orientation of muscle fibres influences the electrical (Zhang et al., 2016) and contractile (Di Natale et al., 2022) activity in smooth muscle organs. Lutton et al. (2017) identified longitudinally and circumferentially oriented fibres in the uterus using groupings of smooth muscle cells. The histological slices can resolve cell nuclei, which allows them to identify individual cells grouped together. With the current settings, μCT could not show that level of cellular detail. However, the quality of the results depends on the resolution of the images. Here, trends in fibre orientations were identified in the uterus using structure tensors that have previously been used in the heart (Rutherford et al., 2012) and pelvis (Yan et al., 2011). The structure tensors identified regions with primarily longitudinal orientation and regions with primarily circumferential orientation, which is consistent with our histology and prior literature. With the higher resolution images of segments of the uterus, regions with primarily longitudinal and circumferential directions were clearly distinguishable. This indicates that the resolution of the images contributes to the quality and efficiency of the structure tensors. However, while structure tensors are able to identify trends in fibre orientation, they do not identify individual fibre structures, nor the exact location of transition between muscle layers. Nevertheless, structure tensors could be used to identify predominant fibre orientations in μCT data, which could be included in multi‐scale models (Zhang et al., 2016) .

The quality of the μCT images depends on the preparation process and the staining agent. The absorption of the stain affects the strength of the signal observed during the imaging process, which can make the distinction between the muscle layers more difficult. Nevertheless, the results presented here showed that μCT is versatile in studying the structures of the uterus. The comparison with the histological slides showed that this method, applied to small segments of the uterine horn, provided similar levels of two‐dimensional information, most notably the possibility to distinguish the longitudinal and circumferential muscles. The major advantage of μCT imaging is the three‐dimensional nature of the information, which is much harder to achieve with classic histology methods. Other 3D approaches using fluorescent microscopy, such as light sheets (Kagami et al., 2020) or confocal microscopy (Jonkman et al., 2020), are also able to produce volumetric data. They rely on specialised pre‐imaging tissue preparation for large samples, such as the rat uterus. This includes refractive index matching to optically clear tissue and the introduction of molecules that preferably bind to targeted proteins and conjugate with a fluorophore. Fluorescent microscopy can be highly discriminatory and enables cell‐scale structural imaging. Thus, confocal microscopy can observe cellular and subcellular structures at high‐resolution. However, μCT is highly effective in this case where the target (the myometrium) is at the tissue level and not the cell scale. The μCT images provided enough information to study fibre architectures in the uterine horns and create an anatomically realistic mesh of the uterus. The present study only looked at the musculature during the oestrus phase of the cycle, yet the uterus undergoes cyclical changes due to the oestrus cycle. Histology has already been used to quantify the changes in the endometrium (Westwood, 2008). This study showed that μCT imaging provided a more complete picture of the rat uterus muscle layers. This method could be used to quantify changes in thickness during the oestrus cycle while simultaneously providing three‐dimensional information about the uterine muscles and fibres.

AUTHOR CONTRIBUTIONS

MR, AG, LC and AC conceived and designed the study. MR, DG and SA collected the data. MR, MT, LC and AC analysed and interpreted the results. MR drafted the manuscript. All authors revised and approved the final version of the manuscript.

FUNDING INFORMATION

This work was funded by the Ministry of Business Innovation and Employment's Catalyst: Strategic fund.

CONFLICT OF INTEREST STATEMENT

The authors have no competing interests to declare.

The code that was implemented for this study can be found on GitHub at https://github.com/virtual‐uterus/uterine‐microCT.

Supporting information

Data S1.

JOA-246-134-s001.zip (1.7MB, zip)

ACKNOWLEDGEMENTS

The authors acknowledge the help of Histology, Anatomy and Medical Imaging, the University of Auckland. The authors thank Linley Nisbet for her technical support. As a partner of the Openness Agreement on Animal Research and Teaching in New Zealand, we acknowledge and appreciate the use of the rat that made this research possible. Open access publishing facilitated by The University of Auckland, as part of the Wiley ‐ The University of Auckland agreement via the Council of Australian University Librarians.

Roesler, M.W. , Garrett, A.S. , Trew, M.L. , Gerneke, D. , Amirapu, S. , Cheng, L.K. et al. (2025) Three‐dimensional virtual histology of the rat uterus musculature using micro‐computed tomography. Journal of Anatomy, 246, 134–147. Available from: 10.1111/joa.14131

DATA AVAILABILITY STATEMENT

The code that was implemented for this study can be found on GitHub at https://github.com/virtual‐uterus/uterine‐microCT S1: µCT and histology datasets https://figshare.com/s/8f190a88930b7367047a (DOI: 10.17608/k6.auckland.23646915) S2: Surface and volumetric meshes generated from the µCT datasets https://figshare.com/s/c21f5dbdc32993388ced (DOI: 10.17608/k6.auckland.23647605) S3: Testing and validation of a structure tensor analysis pipeline for rat uterine µCT images, using synthetic images: https://auckland.figshare.com/articles/dataset/Structure_tensor_validation/25216145 (DOI 10.17608/k6.auckland.25216145).

REFERENCES

  1. Ajayi, A.F. & Akhigbe, R.E. (2020) Staging of the estrous cycle and induction of estrus in experimental rodents: an update. Fertility Research and Practice, 6(1), 1–15. Available from: 10.1186/S40738-020-00074-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Becker, C.M. , Bokor, A. , Heikinheimo, O. , Horne, A. , Jansen, F. , Kiesel, L. et al. (2022) ESHRE guideline: endometriosis. Human Reproduction Open, 2022(2), 1–26. Available from: 10.1093/HROPEN/HOAC009 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bergman, R.A. (1968) Uterine smooth muscle fibers in castrate and estrogen‐treated rats. The Journal of Cell Biology, 36(3), 639–648. Available from: 10.1083/JCB.36.3.639 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Brox, T. , Van Den Boomgaard, R. , Lauze, F. , Van De Weijer, J. , Weickert, J. , Mrázek, P. et al. (2006) Adaptive structure tensors and their applications. Mathematics and Visualization, 0, 17–47. Available from: 10.1007/3-540-31272-2_2/COVER [DOI] [Google Scholar]
  5. Clark, A.R. & Kruger, J.A. (2017) Mathematical modeling of the female reproductive system: from oocyte to delivery. Wiley Interdisciplinary Reviews: Systems Biology and Medicine, 9(1), e1353. Available from: 10.1002/wsbm.1353 [DOI] [PubMed] [Google Scholar]
  6. Di Natale, M.R. , Patten, L. , Molero, J.C. , Stebbing, M.J. , Hunne, B. , Wang, X. et al. (2022) Organisation of the musculature of the rat stomach. Journal of Anatomy, 240(4), 711–723. Available from: 10.1111/JOA.13587 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Dongarra, J. , Koev, P. & Li, X. (2000) Matrix‐vector and matrix‐matrix multiplications. In: Bai, Z. , Demmel, J. , Dongarra, J. , Ruhe, A. & van der Vorst, H. (Eds.) Templates for the solution of algebraic eigenvalue problems: a practical guide. Philadelphia, PA, USA: SIAM, pp. 320–326. [Google Scholar]
  8. Dullin, C. , Ufartes, R. , Larsson, E. , Martin, S. , Lazzarini, M. , Tromba, G. et al. (2017) μCT of ex‐vivo stained mouse hearts and embryos enables a precise match between 3D virtual histology, classical histology and immunochemistry. PLoS One, 12(2), e0170597. Available from: 10.1371/JOURNAL.PONE.0170597 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Farid, H. & Simoncelli, E.P. (2004) Differentiation of discrete multidimensional signals. IEEE Transactions on Image Processing, 13(4), 496–508. Available from: 10.1109/TIP.2004.823819 [DOI] [PubMed] [Google Scholar]
  10. Garfield, R.E. & Maner, W.L. (2007) Physiology and electrical activity of uterine contractions. Seminars in Cell & Developmental Biology, 18, 289–295. Available from: 10.1016/j.semcdb.2007.05.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Garrett, A.S. , Means, S.A. , Roesler, M.W. , Miller, K.J.W. , Cheng, L.K. & Clark, A.R. (2022) Modeling and experimental approaches for elucidating multi‐scale uterine smooth muscle electro‐ and mechano‐physiology: a review. Frontiers in Physiology, 13, 1017649. Available from: 10.3389/fphys.2022.1017649 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Garrett, A.S. , Roesler, M.W. , Athavale, O.N. , Du, P. , Clark, A.R. & Cheng, L.K. (2024) Multichannel mapping of in vivo rat uterine myometrium exhibits both high and low frequency electrical activity in non‐pregnancy. Scientific Reports, 14, 7316. Available from: 10.1038/s41598-024-57734-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Gilbert, S.H. , Sands, G.B. , Legrice, I.J. , Smaill, B.H. , Bernus, O. & Trew, M.L. (2012) A framework for myoarchitecture analysis of high resolution cardiac MRI and comparison with diffusion tensor MRI. Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS, 4063–4066, 4063–4066. Available from: 10.1109/EMBC.2012.6346859 [DOI] [PubMed] [Google Scholar]
  14. Harkness, M.L.R. & Harkness, R.D. (1959) Changes in the physical properties of the uterine cervix of the rat during pregnancy. The Journal of Physiology, 148(3), 524–547. Available from: 10.1113/JPHYSIOL.1959.SP006304 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Jähne, B. (2005) Digital image processing, 6th edition. Berlin, Heidelberg, Germany: Springer‐Verlag. [Google Scholar]
  16. Jain, V. , Saade, G.R. & Garfield, R.E. (2000) Structure and function of the myometrium. Advances in Organ Biology, 8, 215–246. Available from: 10.1016/S1569-2590(00)08009-5 [DOI] [Google Scholar]
  17. Jonkman, J. , Brown, C.M. , Wright, G.D. , Anderson, K.I. & North, A.J. (2020) Tutorial: guidance for quantitative confocal microscopy. Nature Protocols, 15(5), 1585–1611. Available from: 10.1038/S41596-020-0313-9 [DOI] [PubMed] [Google Scholar]
  18. Kagami, K. , Ono, M. , Iizuka, T. , Matsumoto, T. , Hosono, T. , Sekizuka‐Kagami, N. et al. (2020) A novel third mesh‐like myometrial layer connects the longitudinal and circular muscle fibers ‐a potential stratum to coordinate uterine contractions. Scientific Reports, 10(1), 8274. Available from: 10.1038/s41598-020-65299-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Kunz, G. & Leyendecker, G. (2002) Uterine peristaltic activity during the menstrual cycle: characterization, regulation, function and dysfunction. Reproductive Biomedicine Online, 4, 5–9. Available from: 10.1016/S1472-6483(12)60108-4 [DOI] [PubMed] [Google Scholar]
  20. Lammers, W. , Stephen, B. , Al‐Sultan, M.A. , Subramanya, S.B. & Blanks, A.M. (2015) The location of pacemakers in the uteri of pregnant Guinea pigs and rats. American Journal of Physiology. Regulatory, Integrative and Comparative Physiology, 309(11), R1439–R1446. Available from: 10.1152/ajpregu.00187.2015 [DOI] [PubMed] [Google Scholar]
  21. Lutton, E.J. , Lammers, W. , James, S. , van den Berg, H.A. & Blanks, A.M. (2018) Identification of uterine pacemaker regions at the myometrial–placental interface in the rat. Journal of Physiology, 596(14), 2841–2852. Available from: 10.1113/JP275688 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Lutton, E.J. , Lammers, W.J.E.P. , James, S. , Van Den Berg, H.A. & Blanks, A.M. (2017) A computational method for three‐dimensional reconstruction of the microarchitecture of myometrial smooth muscle from histological sections. PLoS One, 12(3), e0173404. Available from: 10.1371/JOURNAL.PONE.0173404 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Malik, M. , Roh, M. & England, S.K. (2021) Uterine contractions in rodent models and humans. Acta Physiologica, 231(4), e13607. Available from: 10.1111/APHA.13607 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Micili, S.C. , Goker, A. , Sayin, O. , Akokay, P. & Ergur, B.U. (2013) The effect of lipoic acid on wound healing in a full thickness uterine injury model in rats. Journal of Molecular Histology, 44(3), 339–345. Available from: 10.1007/S10735-013-9485-8/FIGURES/3 [DOI] [PubMed] [Google Scholar]
  25. Missbach‐Guentner, J. , Pinkert‐Leetsch, D. , Dullin, C. , Ufartes, R. , Hornung, D. , Tampe, B. et al. (2018) 3D virtual histology of murine kidneys–high resolution visualization of pathological alterations by micro computed tomography. Scientific Reports, 8(1), 1–14. Available from: 10.1038/s41598-018-19773-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Moeskops, P. , Wolterink, J.M. , van der Velden, B.H.M. , Gilhuijs, K.G.A. , Leiner, T. , Viergever, M.A. et al. (2016) Deep learning for multi‐task medical image segmentation in multiple modalities. Lecture notes in computer science (including subseries lecture notes in artificial intelligence and lecture notes in bioinformatics), Cham, Switzerland: Springer, vol. 9901, pp. 478–486. Available from: 10.1007/978-3-319-46723-8_55 [DOI] [Google Scholar]
  27. Osa, T. & Fujino, T. (1978) Electrophysiological comparison between the longitudinal and circular muscles of the rat uterus during the estrous cycle and pregnancy. The Japanese Journal of Physiology, 28(2), 197–209. [DOI] [PubMed] [Google Scholar]
  28. Quintarelli, G. , Zito, R. & Cifonelli, J.A. (1971) On Phosphotungstic acid staining. The Journal of Histochemistry and Cytochemistry, 19(11), 641–647. [DOI] [PubMed] [Google Scholar]
  29. Rabotti, C. & Mischi, M. (2015) Propagation of electrical activity in uterine muscle during pregnancy: a review. Acta Physiologica, 213(2), 406–416. Available from: 10.1111/apha.12424 [DOI] [PubMed] [Google Scholar]
  30. Rutherford, S.L. , Trew, M.L. , Sands, G.B. , Legrice, I.J. & Smaill, B.H. (2012) High‐resolution 3‐dimensional reconstruction of the infarct border zone: impact of structural remodeling on electrical activation. Circulation Research, 111(3), 301–311. Available from: 10.1161/CIRCRESAHA.111.260943/-/DC1 [DOI] [PubMed] [Google Scholar]
  31. Testrow, C.P. , Holden, A.V. , Shmygol, A. & Zhang, H. (2018) A computational model of excitation and contraction in uterine myocytes from the pregnant rat. Scientific Reports, 8(1), 1–14. Available from: 10.1038/s41598-018-27069-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Todd, M.E. , Laye, C.G. & Osborne, D.N. (1983) The dimensional characteristics of smooth muscle in rat blood vessels. A computer‐assisted analysis. Circulation Research, 53(3), 319–331. Available from: 10.1161/01.RES.53.3.319 [DOI] [PubMed] [Google Scholar]
  33. Tong, W. , Tribe, R.M. , Smith, R. & Taggart, M.J. (2014) Computational modeling reveals key Contributions of KCNQ and hERG currents to the malleability of uterine action potentials underpinning labor. PLoS One, 9(12), 1–22. Available from: 10.1371/journal.pone.0114034 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Tosti, C. , Pinzauti, S. , Santulli, P. , Chapron, C. & Petraglia, F. (2015) Pathogenetic mechanisms of deep infiltrating endometriosis. Reproductive Sciences, 22(9), 1053–1059. Available from: 10.1177/1933719115592713/METRICS [DOI] [PubMed] [Google Scholar]
  35. Westwood, F.R. (2008) The female rat reproductive cycle: a practical histological guide to staging. Toxicologic Pathology. 36(3), 375–384. Available from: 10.1177/0192623308315665 [DOI] [PubMed] [Google Scholar]
  36. Wray, S. & Noble, K. (2008) Sex hormones and excitation–contraction coupling in the uterus: the effects of oestrous and hormones. Journal of Neuroendocrinology, 20(4), 451–461. Available from: 10.1111/J.1365-2826.2008.01665.X [DOI] [PubMed] [Google Scholar]
  37. Wray, S. & Prendergast, C. (2019) The myometrium: from excitation to contractions and labour. Advances in Experimental Medicine and Biology, 1124, 233–263. Available from: 10.1007/978-981-13-5895-1_10 [DOI] [PubMed] [Google Scholar]
  38. Yan, X. , Kruger, J.A. , Nash, M.P. & Nielsen, P.M.F. (2011) A quantitative description of pelvic floor muscle fibre organisation. In Wittek, A. , Nielsen, P. , Miller, K. (Eds.) Computational Biomechanics for Medicine, New York, NY, USA: Springer, pp.  119–130. Available from: 10.1007/978-1-4419-9619-0_13 [DOI] [Google Scholar]
  39. Zhang, M. , Tidwell, V. , La Rosa, P.S. , Wilson, J.D. , Eswaran, H. & Nehorai, A. (2016) Modeling Magnetomyograms of uterine conttractions during pregnancy using a multiscale forward electromagnetic approach. PLoS One, 11, e0152421. Available from: 10.1371/journal.pone.0152421 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Data S1.

JOA-246-134-s001.zip (1.7MB, zip)

Data Availability Statement

The code that was implemented for this study can be found on GitHub at https://github.com/virtual‐uterus/uterine‐microCT S1: µCT and histology datasets https://figshare.com/s/8f190a88930b7367047a (DOI: 10.17608/k6.auckland.23646915) S2: Surface and volumetric meshes generated from the µCT datasets https://figshare.com/s/c21f5dbdc32993388ced (DOI: 10.17608/k6.auckland.23647605) S3: Testing and validation of a structure tensor analysis pipeline for rat uterine µCT images, using synthetic images: https://auckland.figshare.com/articles/dataset/Structure_tensor_validation/25216145 (DOI 10.17608/k6.auckland.25216145).

Articles from Journal of Anatomy are provided here courtesy of Anatomical Society of Great Britain and Ireland

RESOURCES