Optical Coherence Tomography Enables the Depth-Resolved Measurement of Cilia Beat Frequency in Ex Vivo Human Fallopian Tubes

An example morphological comparison of the distal ampulla on orthogonal views of an OCT image volume and a cross-sectional histology image is shown in Figure 1 . On both modalities, multiple overlapping mucosal folds (plicae) with complex branching are seen. The ciliated columnar epithelium, which is clearly visible on histology, is difficult to discern on static, unprocessed OCT images. In some areas, the epithelium was visible on OCT images as a thin layer contrasting with the underlying lamina propria, and projecting orthogonally from the edges of plicae. The underlying lamina propria, characterized as a fibrovascular core of connective tissue, blood vessels, and lymphatics on histology, was differentiated on OCT by hyperreflective streaks of collagen with hyporeflective gaps. Based on comparison to histology, small gaps in plicae appear to correspond to small blood vessels, while larger gaps appear to correspond to lymphatic channels, but could also be large blood vessels or tangential optical sectioning of cilia-lined folds or branch points in the mucosa. Due to the limited penetration depth of OCT and the thick, complex mucosa of distal ampulla, the underlying smooth muscle layer may not be visualized. The smooth muscle layer was easily imaged on cross sectional OCT images of proximal ampulla segments, which have a simpler planar geometry, less mucosal branching, and a thicker smooth muscle layer. A contrast-enhanced video of a microscopy image sequence is provided in Supplementary Video 1 . On videos, the individual, beating cilia were consistently visualized on the epithelial surface of mucosal folds and grooves, in both focal clusters and continuous spreads. The beat frequency of clusters within the same video FOV was not always uniform and synchronous; some clusters of cilia were observed to beat at distinct speeds, amplitudes and/or directions compared to neighboring clusters. After processing of this video, an example of the CBF color map overlayed on a video snapshot is given in Figure 2A . Image processing quantifies what is seen by eye in the video, that there are differing peak frequencies associated with individual cilia clusters. Line plots of the frequency spectra of three pixels located in example cilia clusters are shown in Figures 2B , 2C , and 2D , showing distinct peak CBF. The overall image sequence CBF calculated was 5.0 ± .23 Hz. Manual review of OCT image sequences clearly revealed beating cilia, visible as a “shimmering” layer with fluctuating intensity lining the edges of plicae. A video of an unprocessed image sequence of a proximal ampulla segment is shown in Supplementary Video 2 . The corresponding quantitative analysis result (CBF color map overlaid on a single OCT cross-sectional image) is shown in Figure 3 . In this proximal ampulla, the ciliated epithelium is limited to a thin hypointense layer overlying a thin hyperintense lamina propria and a thick, visibly textured smooth muscle layer. Nearly all the CBF color map pixels corresponded to pixels with visually fluctuating intensity levels in the OCT image sequence, and hypointense ciliated epithelium. Compared to microscopy, mapped peak frequency values were more variable (heterogeneous) amongst neighboring pixels on OCT color maps, although one prominent cluster had uniform peak frequency of about 2.5 Hz. The peak frequencies of the ciliated epithelium varied between 2–8 Hz, and the overall image sequence CBF calculated was 2.7 ± .23 Hz. As shown in the supplementary material (Figure S1) , spatial smoothing of each B-scan using a 9×9 pixel averaging filter could be used to reduce the apparent heterogeneity without altering the measured CBF value, at the cost of overestimating ciliated areas. Figure 4 shows another example of a quantitative analysis results, in this case a CBF color map overlaid on an OCT image sequence taken from a distal ampulla segment with characteristic overlapping mucosal branches. CBF color map values consistently corresponded to pixels of the ciliated epithelium lining the plicae, even when the epithelium was not visible from the luminal surface. Clusters of distinctly beating cilia are visible, and line plots of the frequency spectra of pixels selected from three clusters are shown in Figure 4B , 4C , and 4D , with peak frequencies of 2.7 Hz, 4.9 Hz, and 7.6 Hz, respectively. Manual review of areas lacking mapped CBF values where cilia were expected revealed that intensity fluctuations visually appeared very slow, or the areas were located in attenuated or out of focus regions. Thus, these pixels were rejected by the minimum peak frequency, frequency amplitude, or intensity amplitude thresholds. These thresholds could be reduced to increase sensitivity to slow moving or attenuated regions, but doing so would introduce a higher number of falsely mapped CBF values. The supplementary material (Figure S2) contains an example of decreasing the minimum signal frequency to 1 Hz, which increases sensitivity but decreases specificity to ciliated epithelium. Potentially, more advanced methods could be developed for more accurate thresholding for CBF mapping. Frequency outliers manifested as variability in waterfall plots (temporal concatenation of line plots of the distribution of mapped pixel peak frequencies), as well as mapping of CBF color values to image pixels beyond the epithelium. An example waterfall plot for image sequences obtained from one patient sample is shown in Figure 5 , whereby each line represents the line plot for each window analyzed. Outliers occurred in at least one window of analysis in 12/37 (32%) image sequences. In five image sequences, variability occurred at low frequencies (2–3 Hz), and corresponded to bulk sample movement (drift) visible in unprocessed image sequences. Sample drift could be either in-plane or out-of-plane of imaging. In 8/37 (22%) of image sequences, variability occurred at high frequencies (5–8 Hz) and was attributed to oscillating sample movement due to transitory equipment vibration. High frequency variability was generally greatest at the start of image acquisition and decayed with time. Figure 6 provides an example of one image sequence, where quantitative analysis of sequential windows reveals initial oscillatory movement (high-frequency outliers), followed by slow drift movement (low frequency outliers), and finally appropriately mapped low-to mid frequency values corresponding to ciliated epithelium. The unprocessed image sequence is provided in Supplementary Video 3 . As shown in Figure 6 , the presence of outliers in non-ciliated regions did not appear to affect measurement of peak frequency values in the appropriately mapped, expected ciliated regions. Two image sequences contained outliers which persisted over the entire duration of imaging, and were excluded from further quantitative analysis. On all other image sequences, defining the overall image sequence CBF as the mode (the frequency bin with the maximum value) was robust to outliers, and our total imaging time (~ 9 s) appears to be sufficient to measure CBF even in the presence of high and low frequency outliers. A summary of the characteristics and mean tissue CBFs calculated for each sample is shown in Table 1 . Four patients were premenopausal, and one was postmenopausal. OCT images were acquired within 2.5 hours of explant. Sample temperatures measured between 17.5–20°C at the start of OCT image acquisition. Between 5 and 10 sequential OCT image sequences were obtained on each patient’s sample (n = 37 image sequences total). For each patient, CBF measurements were similar across OCT image sequences acquired from different areas of the sample. CBF measured from OCT image sequences across all patients ranged from 2.2 to 3.6 Hz. For all but one tissue sample, the corresponding CBF values measured from video microscopy image sequences differed by 1.4 Hz or less.

One ampulla FT segment of 1–2 cm length was obtained from 5 females following laparoscopic hysterosalpingectomy for benign conditions. Informed consent was obtained verbally from each patient under an Institutional Review Board protocol approved on November 14, 2024 by the University of Arizona (1100000679). The indications for surgery included abnormal uterine bleeding (n = 2) and pelvic organ prolapse (n = 3). The explant time, defined as removal of the specimen from the body, was noted. The sample was transported in cold (~5°C) Dulbecco’s modified Eagle’s medium (DMEM). The ampulla was longitudinally opened to expose the lumen and rinsed with room temperature phosphate-buffered saline (PBS). Excess connective tissue was carefully resected and the sample was pinned to a transparent silicon-coated petri dish to expose the luminal surface. The sample was submerged in room temperature PBS. Imaging duration was calculated as the time between surgical explant and the end of OCT image acquisition. Cilia movement was validated using an upright brightfield microscope (BX41, Olympus, Japan), with oblique illumination. Videos of at least two areas with resolvable cilia were acquired using a 50x magnification objective and 4.0-megapixel resolution camera (ORCA Flash 4.0 V2, model C11440 –22CU). The field of view (FOV) was set to 1024 × 1024 pixels (approximately 210 × 210 μm), corresponding to a pixel size of approximately 0.2 μm. Sequential images were obtained at a frame rate of 60 frames per second (fps). Image sequences were cropped to exclude out of focus or non-ciliated portions of the images, and the image was contrast enhanced to define ciliary clusters, using ImageJ (NIH, Bethesda, MD). Immediately after acquisition of video microscopy images, the sample was transferred to the OCT imaging stage. Sample temperature was measured at the start of OCT image acquisition. The sample was imaged using a commercial benchtop OCT system (Thorlabs Telesto Tel221PS) and corresponding software (ThorImageOCT 5.6.0), with an axial resolution of 11 μm in air and center wavelength of 1300 nm. A 10x magnification objective with a lateral resolution of 7 μm was used (Thorlabs LSM02). In 5–10 different regions of the sample, at least 500 sequential B-scans, or two dimensional (2D) depth resolved scans, were acquired at a frame rate of 57 fps and an A-scan rate of 76 kHz. The frame rate was confirmed using an oscilloscope. The FOV was 2 mm (lateral) × 2.53 mm (axial), corresponding to 1000 × 1024 pixels, which translates into a pixel size of 2 μm × 2.47 μm. Using the same spatial sampling, a 3-dimensional (3D) scan at least 2 mm (lateral) × 1 mm (lateral) × 2.53 mm (axial) of the sample was acquired, to characterize the structural morphology. Following image acquisition, samples were fixed in formalin and processed for histology using routine techniques. Hematoxylin and eosin (H&E) stained sections were reviewed for general feature morphology (e.g. plicae structure and presence of cilia), and the morphology was qualitatively compared to 3D OCT images. Unprocessed OCT and microscopy image sequences were qualitatively reviewed using ImageJ. Structural characteristics as well as any image artifacts or sample movement were noted. The highly reflective air-PBS boundary seen in OCT images was manually cropped out of images to avoid false mapping of CBF values 16 . Spatially-resolved CBF was computed using an algorithm previously reported 14 – 17 and implemented in MATLAB (2023b). Briefly, for each image sequence, a manually determined grayscale intensity threshold (in our case 60, out of 0 to 255 8-bit images) was universally applied to exclude processing of pixels in hyporeflective (non-tissue) areas, such as media, background noise, or regions with attenuated signal. Fast Fourier Transform (FFT) with a sliding-window in time was performed at each supra threshold pixel, utilizing 128 images (~2.2 s) of data in each window and a window increment of 1 s, to acquire the frequency spectrum with a resolution of ~0.45 Hz. After FFT, an amplitude threshold, defined as the mean frequency amplitude plus 3 standard deviations in the frequency range, where cilia beat is not expected (>23 Hz), was calculated. Only signals within the expected frequency range of CBF (2–10 Hz) with amplitudes above this threshold were analyzed. The normalized frequency spectrum between 2 and 10 Hz was measured for each pixel, and the frequency with the peak amplitude identified. The “peak frequency” was color coded, and a CBF color map image was created by spatially mapping the peak frequency of each pixel that met both the grayscale intensity and the frequency amplitude thresholds. CBF color maps were overlayed onto the intensity image to reveal the location and peak frequency of cilia. Additionally, a line plot was generated to visualize the distribution of mapped pixel peak frequencies. This process was repeated for each sliding window. A concatenation of line plots (a waterfall plot) was generated to visualize if and how frequency content changed over time. A column chart histogram was created to summarize the normalized pixel frequency distribution across all windows of the image sequence. Finally, the image sequence CBF was defined as the frequency bin with the maximum value (the mode), presented as the bin center ± bin width/2 (Hz). CBF color map overlays were qualitatively compared to unprocessed OCT and microscopy image sequences. Mapped CBF values were expected only in areas corresponding to the ciliated epithelium (i.e. the edges of plicae), since movement in this frequency range is not expected to occur in any other part of the FT (peristalsis of smooth muscle should be much lower frequency). Image sequences which showed temporal variability in waterfall plots, or had spatially mapped CBF values in unexpected areas, were further reviewed to identify potential sources of artifacts and noise. For each sample, the CBF across image sequences acquired from different sample locations was calculated and reported as the mean tissue CBF separately for OCT and video microscopy imaging (mean ± standard deviation [Hz]). Descriptive characteristics for each patient, as well as CBF measurements using video microscopy, were reported but were not statistically compared due to the small sample size and lack of spatial and temporal co-registration between imaging methods.

For the first time, CBF was measured and spatially mapped in human fallopian tubes, from OCT image sequences. Despite the inability to resolve individual cilia, their movement was manifest as visible fluctuations in intensity on OCT images. CBF color maps and histograms provided insights into cilia dynamics present in multiple overlapping layers of fallopian tube mucosa, which vary in complexity along the length of the fallopian tubes. Frequency outliers in some images occurred due to sample or instrument movement, but were typically not present for the entire duration of imaging, and therefore could be calculated from 9 seconds of acquisition. Our results collectively indicate that OCT is a feasible approach to structurally and functionally characterize cilia dynamics within the fallopian tubes in the ex vivo setting. Future work will be directed toward optimizing our experimental protocol to reflect in vivo conditions, as well as evaluating natural hormonal and pathological factors which may impact the coordinated movement of cilia.

We demonstrate for the first time, the feasibility of using OCT Fourier analysis to reveal the location and frequency of cilia beating in the human fallopian tubes. Despite the inability of current OCT systems to resolve individual cilia, we showed that their movement was clearly visible in unprocessed images as fluctuations in pixel intensity over time. Using an algorithm previously developed for imaging mouse oviduct, we showed that these intensity fluctuations encode cilia movement and can be used to compute cilia beat frequency. Our method has the potential to provide insights into fundamental aspects of reproductive physiology and may also improve our current understanding of pathogenic mechanisms or markers of tubal factor infertility, the most common cause of female infertility 11 . This approach, coupled with the recent development of OCT falloposcopes 19 , may ultimately pave the way for a new approach to non-destructive, minimally-invasive and potentially office-based assessment of the fallopian tube cilia function in vivo . The morphology and geometry of mucosal folds, as well as the density and distribution of cilia, are complex in humans and vary along the length of the fallopian tubes and over time 3 , 4 . Compared to traditional high-speed video microscopy, OCT provides a large, depth-resolved field of view of the endotubal mucosa, enabling the assessment of underlying structures in addition to functional surveillance of ciliated areas hidden beneath overlapping plicae. In this study, we observed both continuous distributions and clusters of cilia beating at distinct frequencies on both OCT and microscopy images. The scale and inclusion of multiple cilia per pixel may account for the observed heterogeneity in mapped CBF values on OCT images, and may reflect natural variations in velocity or focal differences in cilia coverage that were observed on high-magnification microscopy images 4 , 17 , 22 , 23 . The indirect measurement of CBF based on speckle fluctuations also makes OCT more susceptible to noise introduced by non-periodic motion (e.g., red blood cell movement, media particulates, sample drift) and other tissue scatterers (e.g., mucus) 23 , 24 . Despite this limitation, our results may provide insight into conflicting and variable CBF measurements reported in prior studies, which have been limited to a small FOV of superficial areas or have lacked imaging completely 3 . For each patient sample, our CBF measurements were generally consistent over approximately 9 seconds of imaging time. Our image frame rate of 57 fps enabled a highest detectable frequency of 28.5 Hz, which is sufficient to detect human CBF at room temperature (expected up to 25 Hz 3 ). The inclusion of 128 frames (2.25s) for analysis provided a balance between frequency resolution (0.45 Hz) and sample movement, an inherent limitation of longer imaging time. Obtaining longer image sequences and using a sliding window approach helped assure that a frequency outlier-free window would be obtained, and assisted in identification of the cause of outliers. In future work, we will investigate motion stabilization and noise filtering strategies to mitigate erroneous measurement due to patient movement or operator instrumentation handling, which is even more likely to occur in the clinical setting 12 . Also, in an expected clinical implementation, CBF color maps would be processed and displayed in real time, and trained operators could retake images if unexpected outliers were present (similar to the operation of Doppler ultrasound). Our measurements of CBF from OCT images consistently ranged between 2.2 – 3.6 Hz for all human ampulla samples, which is slower than that reported in vivo (~5 Hz) 12 and ex vivo using microscopy (3.4 – 8.7 Hz) 3 , 25 . In this study, OCT-derived CBF measurements tended to slightly underestimate those measured by video microscopy; however, the differences were within biologically insignificant limits (<1.5 Hz) 24 in all but one tissue sample. Cilia movement is tightly regulated, and several factors may contribute to this discrepancy. First, the ex vivo environment in this study included sub-physiological temperatures (<20°C) 14 , 26 , a nutrient-depleted tissue environment (PBS media), and disruption of perfusion and progressive hypoxia following electrosurgical removal of the tissue. Small differences in temperature (~1 Hz/°C expected 24 ) and the longer delay in imaging time between OCT and microscopy could in part, explain the difference in measured values between modalities. Second, unlike traditional studies that quantify CBF by manually selecting small, visibly beating regions—typically at the level of a single pixel or cilia cluster—our OCT method performs automated, pixel-wise analysis across a large field of view. A potential limitation of this approach is the inclusion of regions where cilia are denuded, physically obstructed (e.g., by mucus), or absent altogether, which may lower the detected frequency 23 , 24 . Finally, CBF depends on patient-specific factors including age, hormonal status 9 , and concurrent pathology (even if pathology is not evident on tubal histology) 3 , which were not formally considered in this study. Future ex vivo studies will more tightly control imaging variables and examine cilia dynamics differences due to hormonal and pathological factors, including changes across the menstrual cycle or functional outcomes of disease linked to infertility (such as endometriosis). Additional limitations of this feasibility study include the small sample size (n=5) and difficulty co-registering OCT, video microscopy, and histology images. Agreement between spatially co-registered CBF measurements on OCT and video microscopy using the same analysis method employed herein was previously validated in the mouse oviduct in vitro 17 . It is also important to note that, while comparative ex vivo studies have reported reduced CBF in the setting of infertility-causing pathologies 4 – 6 , 8 , 10 , the clinical significance of CBF values alone remains uncertain and may not comprehensively reflect the functional status of the human fallopian tubes in vivo . Of greater clinical interest is to investigate large-scale dynamic interactions between neighboring cilia, which generate metachronal waves to directionally propagate luminal contents such as embryos and fluid 14 , 27 . In future ex vivo work, we will implement location-matched high-speed video microscopy to validate our qualitative observations, and utilize OCT to characterize the phase, direction and velocity of metachronal waves to further investigate how impairments in cilia-induced flow patterns may contribute to infertility.

The fallopian tubes (FTs) provide an ideal environment for the complex and tightly regulated processes of natural conception 1 . The periodic beating of small (4–7 μm long), hair-like projections lining the endotubal lumen, called cilia, plays an essential role in oocyte pickup, embryo and sperm transport, and clearance of luminal debris and waste 2 , 3 . Impaired cilia motility is associated with serious clinical complications and subfertility in women, and is linked to several pathologies including endometriosis 4 – 6 , adenomyosis 7 , pelvic inflammatory disease 8 , ectopic pregnancy 9 , and obstruction 10 . Despite the recognized importance of FT motile cilia, their physiological function in humans and their role in tubal factor infertility, which affects approximately one-third of infertile women 11 , remain understudied. The ability to quantify cilia movement in situ has great potential to improve understanding of the role of normal and impaired cilia movement in human reproduction, and inform future approaches to diagnosis and management of tubal pathology and subfertility. The periodic movement of cilia, quantified as cilia beat frequency (CBF), has been used to characterize normal FT function and identify impairments in the setting of tubal disease ex vivo using high-speed video microscopy and other non-imaging light-scanning methods 3 . Published values are highly variable 3 , which is in part due to the lack of standardized tissue preparation and measurement protocols, and inherent limitations surrounding removal of tissue from its natural microenvironment. Paltieli et al are the only group to report CBF in humans in vivo , using a laser scanning instrument to measure fluctuations in light scattering 12 . Their approach was invasive (laparoscopic), could only measure the distal fimbriae segment, and did not provide spatial information. The development of a minimally invasive method to image and quantify CBF within the FTs has great potential to serve as a useful diagnostic and screening tool for women with low-grade tubal disease or unexplained infertility. A novel algorithm was previously developed to quantify and spatially map cilia dynamics in live mouse oviducts, based on quantifying intensity fluctuations in optical coherence tomography (OCT) images 13 – 17 . OCT is an advantageous approach to study FT dynamics because it provides rapid, depth resolved images, without the need for exogeneous contrast agents 18 . Recent developments have enabled implementation of OCT into miniature (<1 mm) multi-modal fiber optic probes to assess FT microstructure in vivo 19 , 20 and ex vivo 21 . These developments provided direct OCT imaging access to the FT ciliated epithelium. In the present study, we demonstrate the feasibility of OCT imaging and analysis to evaluate human FT cilia dynamics in ex vivo tissue samples, as a first step toward clinical translation.