1.1.

Origin and Characteristics of Early Stage Ovarian Cancer

Due to nonspecific clinical symptoms and a lack of effective screening methods, it is exceedingly difficult to diagnose ovarian carcinoma in its early stages.^1,2 Because of the low sensitivity and/or positive predictive value of bimanual ovarian palpation, carbohydrate antigen 125 (CA125) blood tests, and transvaginal ultrasound, even when performed in combination, screening for ovarian cancer can require invasive surgery and removal of the ovaries to make a definitive diagnosis.^3–7 As a result, 79% of cases are not discovered until they are already at an advanced stage when five-year survival rates are as low as 29%.^1,8,9 A large number of risk factors for developing epithelial ovarian cancers have been identified, including a family or personal history of breast or ovarian cancer, gene alterations (e.g., BRCA1/2), reproductive history, and advanced age.^8,10–13 Patients at known risk may be counseled to undergo a prophylactic bilateral salpingo-oophorectomy, which reduces ovarian cancer risk by up to 81%, but results in immediate onset symptoms of postmenopause and infertility for premenopausal patients, as well as an increased risk of osteoporosis and cardiovascular mortality long-term.^14–16

Mounting evidence suggests that most cases of high-grade serous ovarian carcinoma originate as serous tubal intraepithelial carcinoma (STIC) lesions in the fallopian tube epithelium (FTE), rather than lesions in the ovarian surface epithelium.^17–20 STIC lesions, whether invasive or noninvasive, may spread through exfoliation onto the surface of the ovary or into the peritoneal cavity.²¹ Modeling of the growth and progression of these lesions also suggests that there may be a window of detection lasting several years before they metastasize and advance to ovarian cancer.^22,23 These findings suggest that a device capable of detecting STIC lesions in the fallopian tubes may enable early stage detection of ovarian cancer. For cancer caught at this stage in the fallopian tube, five-year survival rates may be 95% or greater, and treatment may require neither radical surgery nor adjuvant chemotherapy.^1,6,21

STIC lesions present as clusters of cytologically malignant epithelial cells up to hundreds of microns in diameter that replace normal fallopian tube mucosa, most commonly, but not always, involving the fimbria.^24,25 These lesions are smaller than what can typically be seen by whole-body (magnetic resonance imaging, computed tomography, or positron emission tomography) or ultrasound imaging modalities.⁷ This small size of STIC suggests that any visualization method must have a resolution of tens of micron. Additionally, the method should ideally be non-ionizing and comprehensively visualize the entire 11- to 12-cm length of the fallopian tube, as STIC may occur at any location.^7,22,24–26

Generalizable hallmarks of malignant transformation include the breakdown of collagen and remodeling of tissue, increased cellular metabolism, and increased vascularization.^22,24,25 Therefore, the optical (reflectance and autofluorescence) signature of STIC lesions is expected to be altered.^22,27 Changes in the abundance and accessibility of cell surface receptors and extracellular proteins also occur in STIC and subsequent ovarian cancer, including folate receptors and laminin-C1 proteins, which may be targeted with fluorescence-tagged ligands.^28,29 STIC lesions are also characterized by unique cellular and molecular features, and thus could be identified with the use of various cytological or omics analyses.^17,18,24,25 These characteristics indicate that a combination of reflectance imaging, fluorescence imaging, and guided biopsy of epithelial cells from suspect lesions in the fallopian tube may be used to identify STIC before it metastasizes to the ovarian surface epithelium. Such a screening method, designed for the detection of STIC, may also be widely applicable to the detection of other pathology and tubal disease that presents in the FTE or tubal lumen.

1.2.

Methods for Early Detection of Ovarian Cancer

Any early detection method should be minimally invasive and have high positive predictive value (PPV). The most promising results to date come from serial measurement of CA125 levels and application of the risk of ovarian cancer algorithm (ROCA), followed by transvaginal ultrasound in women at high risk.⁵ Still, the PPV with ROCA is only 10.8% for ovarian or fallopian tube cancers.³⁰ Optical imaging is a relatively inexpensive, non-ionizing alternative to whole-body imaging and ultrasound that can detect either endogenous changes in reflectance and autofluorescence or the presence of contrast agents and fluorophores targeted to disease-related biomarkers.^{22,27,28,31–33} An optical system using reflectance and fluorescence imaging achieved high sensitivity, specificity, and negative predictive value for the detection of clinically occult and cancer-related lesions within ex vivo fallopian tube tissue. However, it had high false-positive rates and therefore low PPV.²² These results suggest that reflectance and autofluorescence may be best utilized for identification of suspicious areas, to be followed by confirmation of disease with tissue sample acquisition for pathological, cytological, or omics-based analyses.^28,34

The major drawback of optical imaging, as with biopsy, is the need to gain physical access to the tissue of interest. Because the uterotubal ostium has a diameter $<1\mathrm{mm}$ across at its narrowest point, size constraints have previously prevented the inclusion of both optical imaging and biopsy in a single endoscopic system capable of operating within the tubal lumen without surgical access.^26,35–39 However, advances in optics and photonics have led to the progressive miniaturization of fibers, lenses, light sources, detectors, and other imaging components, which has resulted in new access to narrow anatomical channels for multifunctional optical endoscopes. Simple optical endoscopes capable of navigating the fallopian tube lumen have been demonstrated in vivo, but they have primarily been used for the assessment of tubal patency and identification of stenoses with white-light imaging.^36–40 They have had outer diameters as small as 0.50 mm and captured images through a 2000- to 3000-element fiber bundle with optics to obtain up to $50\times$ magnification.^36–39 To distend the fallopian tube for improved imaging performance, these devices typically used either a liquid flush, or a linear-everting balloon.^35,39 Because these falloposcopes often encountered challenges such as intense light reflections, lenses soiled by floating debris, and insufficient lumen distension, their use in routine clinical practice remains limited.^35,39,41 Still, these simple falloposcopes could sometimes identify tubal causes of infertility.^35,39,41 Recently, a 1.2-mm-diameter scanning-fiber endoscope was utilized for high-resolution white-light imaging of the fallopian tube.⁴² The endoscope had difficulty entering and traversing the length of the tube, possibly due to the relatively large diameter of the endoscope.

Advancement of manufacturing technology, miniaturization of mechanical components, and the development of new, flexible, sterilizable, and biocompatible materials has allowed for the construction of devices such as the Cytuity™ catheter, which is capable of in vivo cell collection from the entire proximal portion of the fallopian tube.⁴³ However, this device does not include imaging nor enable directed sampling of abnormal tissue. Thus, flexible, steerable biopsy devices, and multimodal optical endoscopes exist that are capable of operation within the fallopian tube lumen, yet no falloposcopic device has included imaging and biopsy together for the comprehensive detection of STIC and other pathology in the FTE.

Here, we present the design, construction, and feasibility testing of a flexible and steerable submillimeter-diameter endoscope capable of multispectral reflectance imaging, multispectral fluorescence imaging, and guided cell collection in the fallopian tubes. We describe the design and construction of our falloposcope based on clinical requirements for detecting STIC, detail the experimental testing of the optical imaging and cell collection subsystems, and summarize the results, which demonstrate that the device is capable of capturing images of adequate resolution for detecting STIC, as well as collecting a sufficient number of cells from the FTE for relevant downstream analyses and diagnosis.

2.1.

Clinical Requirements and System Specifications

Table 1 outlines the clinical requirements and technical specifications for a falloposcope capable of imaging and cell collection. Although these requirements were primarily developed with the goal of detecting STIC, they are also applicable to the detection of other tubal pathology including salpingitis, tubal occlusions, polyps, tumors, and other endotubal lesions or disease.^44,45

Table 1

Requirements for the design of a minimally invasive, optical endoscope capable of both imaging and cell collection for in vivo detection of STIC within the FTE.

2.2.

Operational Procedure

The cell-acquiring fallopian endoscope (CAFE) was designed to meet all of the requirements outlined in Table 1 and be compatible with as much low-cost, standard instrumentation as possible, making it operationally straightforward for use in an outpatient procedure. The endoscope is small enough to be inserted into the uterus through a standard hysteroscope and directed to the uterotubal ostium using a standard salpingography introducing catheter. A hemostasis valve Y-connector placed on the introducing catheter allows for the delivery of liquid or gas as needed to distend the fallopian tube around the CAFE. The CAFE can be navigated through the tubal lumen and used to locate sites of potentially cancerous tissue, as indicated by changes in the optical signature of the tissue. These signs include an alteration in natural tissue reflectance or autofluorescence, or an increase in targeted fluorescence within the FTE.^22,27,28,31 A curved cell collection wire can be extended from the distal end of the CAFE to obtain a directed scrape biopsy of cells from regions of interest, and then retracted into the endoscope as it is withdrawn from the patient. After the conclusion of the procedure, the cells can be dissociated from the wire to yield material usable for confirming visual predictions with karyometry, proteomics, and other cytological or omics analyses used for the diagnosis of cancer and other pathology.^{17,18,24,25,34} This procedure is shown in Fig. 1.

Fig. 1

The CAFE is inserted into the uterus and navigated through the fallopian tubes to locate sites of potentially pathological tissue. A scrape biopsy of cells from the region of interest is obtained with the cell collection wire, and the cells can be used for further analysis.

2.3.

Falloposcope System Design

The CAFE falloposcope system consists of a single-use, handheld endoscope that is inserted into the patient, and a reusable, portable console that contains all of the powered components for illumination and imaging. Connection between the console and endoscope occurs at the top surface of the console inside a protective connection box. The CAFE is designed to be inserted through a standard hysteroscope and introducing catheter. Photographs of the completed prototype system are shown in Fig. 2.

Fig. 2

(a) Sterilization case and prototype handheld endoscope inserted into the introducing catheter and hysteroscope. (b) Distal tip of the prototype endoscope with the cell collection wire extended. (c) Full prototype system.

2.3.1.

Console

The console is a modified, portable equipment rack [Middle Atlantic Products, part number (PN): PTRK-21, Fairfield, New Jersey], containing all power management, computing, illumination, and detection components. A block diagram of these items is shown in Fig. 3. A power isolator [Tripp Lite, PN: IS1000HG 100W Isolation Transformer, Chicago, Illinois] assures conformity with electrical leakage considerations. A computer (HP, Z2 Mini G4 Workstation, Palo Alto, California) with a monitor, keyboard, and mouse enables control of the imaging components, real-time visualization through the endoscope, and storage of captured images and videos. The console houses the illumination sources for imaging, which are four fiber-coupled laser diode sources (Thorlabs, PN: LP642-SF20, LP520-SF15, LP488-SF20, and LP405-SF10, Trenton, New Jersey) operating at emission wavelengths of 405, 488, 520, and 642 nm. These wavelengths are useful for distinguishing the visual features of healthy and diseased tissue with reflectance and fluorescence imaging, and they are readily available as commercial, off-the-shelf components.^22,27,28,31 The four diodes can be selectively turned on and off for reflectance and fluorescence imaging at each individual wavelength, or multiplexed through a custom, fiber-coupled wavelength division multiplexer [OZ Optics, Ontario, Canada] for imaging with any combination of the four sources. Currently, each laser source is controlled and switched manually using the touchscreen interface of its laser diode driver [Thorlabs, PN: CLD1010LP], but this could be automated in the future. Laser illumination is carried to the endoscope tip by a single illumination fiber. Narrowband reflectance imaging can be performed to increase the contrast of all vasculature in the tissue, whereas fluorescence imaging can be used to detect either targeted fluorophores or metabolic and microstructural changes in the tissue.²² Pseudo-white-light imaging can be accomplished by overlaying reflectance images taken sequentially at each wavelength. The total laser power incident upon the tissue must comply with ANSI standard Z136.1. For continuous illumination with visible wavelengths, radiant exposure is limited to $0.2\mathrm{W}/{\mathrm{cm}}^2$, corresponding to about 3.2-mW total laser power emitted from the tip of the endoscope at a working distance of 5 mm from the tissue.

Fig. 3

Block diagram illustrating the contents of the CAFE console, responsible for power, control, illumination, and imaging, represented in sections from bottom to top. Components shown in light yellow make up the illumination and imaging subsystems.

The image captured by the handheld endoscope is relayed to the console through a fiber bundle. The end of the fiber bundle is imaged by a $40\times$ plan achromatic microscope objective [Olympus, PN: PLN-$40\times$, Center Valley, Pennsylvania] with a 0.65 NA, and the image is passed through a fold mirror, filter wheel [Thorlabs, PN: ELL9K], and an achromatic lens [Melles Griot, PN: 01LA0119, Rochester, New York] to focus on a six-megapixel monochrome CCD camera [Retiga, PN: R6, Surrey, BC, Canada] with $4.54\mu \mathrm{m}$ pixels, thermoelectric cooling, and a 75% peak quantum efficiency. The filter wheel positions an appropriate long-pass filter in front of the camera for fluorescence imaging with each laser diode. This includes a 450-nm long-pass filter [Thorlabs, PN: FEL0450] for the 405-nm laser source, a 500-nm long-pass filter [Thorlabs, PN: FEL0500] for the 488-nm laser source, a 550-nm long-pass filter [Thorlabs, PN: FEL0550] for the 520-nm laser source, and a 650-nm long-pass filter [Thorlabs FEL0650] for the 642-nm laser source. One slot in the filter wheel is left empty for reflectance imaging. Ocular software [Teledyne Photometrics, Ocular Image Acquisition Software, Tucson, Arizona] is used to set exposure time, adjust image settings (brightness, contrast, etc.), capture and save still images, record and save video, add false color, and overlay images for pseudo-white-light image capture and navigation through the fallopian tube. Image acquisition speeds of $\sim 10$ frames per second can be achieved with a 0.100-s exposure time. Separate software [Thorlabs, Elliptec Systems Software] controls the filter wheel and maintains a 450- to 500-ms switching time.

2.3.2.

Endoscope

The handheld endoscope utilizes many of the same design principles as the single-use, multimodal falloposcope described in a previous publication from our research group.⁴⁰ However, the endoscope previously described did not have cell collection capability. The CAFE has multiple components packed into a 70-cm-long insertable body that is 0.79 mm in diameter and is connected to a handle with unidirectional steering control. The main body is a flexible, 0.71-mm-diameter multi-lumen extrusion (MLE) of polyetheretherketone (PEEK) plastic, which contains four channels of different sizes, shown and labeled in Fig. 4. The image collection channel of the MLE is a circular, 0.30-mm-diameter channel containing a 3000-element image fiber bundle [Fujikura, PN: FIGH-03-200S, Tokyo, Japan] with an attached gradient index (GRIN) lens [GRINtech, PN: GT-IFRL-025-005-50-CC, Jena, Germany]. On each side of the MLE, there are circular channels 0.13 mm in diameter: the pull-wire channel houses a 0.10-mm-diameter, 304 V stainless alloy pull-wire [Fort Wayne Metals, Fort Wayne, IN], whereas the illumination channel houses a 0.10-mm-diameter, teflon-clad, multimode illumination fiber with an $85\mu \mathrm{m}$ core diameter and a 0.66 NA [Polymicro Technologies, PN: FSU085100, Phoenix, Arizona]. The pull-wire enables up to 60-deg unidirectional steering at up to a 25-mm bend radius (bend limit of the image collection fiber bundle).

Fig. 4

(a) Exploded view of all handheld endoscope components contained within the MLE at its distal tip. (b) Rendered image of the distal tip of the endoscope with the cell collection wire extended and the MLE channels labeled.

The cell collection channel is 0.26-by-0.57 mm in diameter and has a gibbous shape for housing the curved cell collection wire. This channel, which occupies all remaining space in the endoscope, is too small to accommodate commercially available biopsy devices, baskets, or cytology brushes. A small diameter nitinol wire can be shape-memory formed such that can traverse the small diameter channel, and then assume a gentle curve when extended out the distal end of the endoscope. Various diameters and shapes of wire can be used.

The rest of the handheld endoscope design closely follows the design of the multimodal falloposcope for clinical use detailed in our research group’s previous publication.⁴⁰ In summary, multiple sections of optically transparent, thin-walled, medical grade heat shrink tubing [Nordson Medical Corp., PN: 103-0139, Westlake, Ohio] protect the distal tip of the endoscope, bringing the total outer diameter of the insertable portion of the endoscope to 0.79 mm. Multiple layers of thicker heat shrink [Nordson Medical Corp., PN: 103-0454] tubing are used on the rest of the PEEK MLE to create a biocompatible, smooth, sterilizable surface with an overall diameter of 0.79 mm. A 15.25-cm-long, 1.87-mm-diameter stainless steel hypotube [Vita Needle, Needham, Massachusetts] supports the MLE where it attaches to the endoscope handle.

All endoscope components run through an 18-cm-long, 3D-printed, ergonomic handle. Renderings of the handle with all internal components labeled are shown in Fig. 5. The stainless steel pull-wire wraps around a pulley that is attached to a steering wheel on the outside of the handle, enabling unilateral tip deflection. The entire endoscope can be rotated to control the direction of deflection. The cell collection wire runs through a separate channel in the handle that is lined with lubricious tubing and mated to the MLE cell collection channel. A female Luer Lock adapter at the surface of the handle can be used for the delivery of liquid through the cell collection channel when the cell collection wire is not inserted, enabling the clearing of debris from the endoscope optics and dispensing of exogenous fluorophores into the fallopian tube. Maneuverability and control of the cell collection wire by the operator is improved by a wire torquer attachment attached near the proximal end of the wire. The image collection fiber bundle passes through a straight, central channel in the handle, whereas the illumination fiber follows a separate, lateral channel that contains a custom mode scrambler for adjusting the output angle of illumination from the endoscope.⁴⁰ A 3-m long, vinyl-encased cable connects the endoscope to the console.

Fig. 5

(a) Image of the prototype handheld endoscope. (b) Image of the opened endoscope handle showing all internal components. (c) Rendered image of the components inside of one half of the handheld endoscope. (d) Rendered image of the handheld endoscope with all internal components labeled.

2.4.

Experimental Setup and Characterization

A prototype CAFE console and endoscope were constructed and tested to assure they met the requirements for imaging, illumination, and cell collection outlined in Table 1.

2.4.1.

Imaging and illumination testing

As shown in Fig. 6, a 3D-printed, fixed-geometry jig was used for repeatable testing of imaging and illumination in a dark environment. The jig holds the endoscope tip 5 mm from a resolution standard, and 6.75 mm from a 1-cm square cuvette containing a fluorophore. For resolution measurements, the 488-, 520-, and 642-nm laser diodes were turned on simultaneously and images of a standard 1951 USAF Resolution Target were captured. The images were analyzed in ImageJ [NIH, Version 1.8.0_112], defining resolvable elements as those with a minimum modulation contrast of 15% for both horizontal and vertical bars, based on the Rayleigh criterion.⁴⁶ Field of view was measured by imaging a one-line-pair-per-millimeter (lp/mm) target. In ImageJ, the width of one line pair was used to set the scale, and then the distance over the entire imaging circle was measured.

Fig. 6

(a) Rendered image of the illumination and imaging test setup with a 1951 USAF Resolution Target and cuvette of liquid fluorescence standard placed in the 3D-printed jig. (b) Cross-section of the test setup with endoscope inserted and components labeled.

Illumination and fluorescence imaging sensitivity were tested using cuvettes of fluorescent dyes—acridine orange ($18\text{-}\mu \mathrm{M}$ dissolved in ethanol), fluorescein ($2\text{-}\mu \mathrm{M}$ dissolved in 0.1 M sodium hydroxide), and rhodamine ($0.8\text{-}\mu \mathrm{M}$ dissolved in ethanol) – which produce remitted fluorescence signal similar to what is expected from tissue at excitation wavelengths of 405, 488, and 520 nm, respectively. An image was captured with a 100-ms exposure time through the appropriate fluorescence filter. Reflectance and fluorescence images were obtained with and without laser illumination to measure illumination uniformity and signal-to-background ratio. Qualitative analysis of image quality was assessed from reflectance (monochrome, narrow band, and pseudo-white-light) and fluorescence images that were acquired in fresh ex vivo porcine fallopian tube tissue. Reflectance and fluorescence images were also captured in ex vivo human fallopian tube tissue, but these image acquisition tests were used to assess the feasibility of fitting the endoscope into human fallopian tubes for imaging, rather than generalizable formal analysis of image quality. The human tissue was unsuited for image analysis because all samples acquired as surgical discard from procedures involving salpingectomy were limited to $<1\mathrm{cm}$ in length through our Institutional Review Board human subjects protocol with the Banner University Medical Center Department of Pathology.

3.1.

Imaging and Illumination

All imaging requirements in Table 1 were met. Analysis of the images of the 1951 USAF Resolution Target [as shown in Fig. 7(a)] showed that the smallest resolvable element was group 2 element 4 when acquired from a distance of 5 mm from the target. This corresponds to a resolution of $88.4\mu \mathrm{m}$. The full field of view was $\sim 45\mathrm{deg}$ in air. Fluorescence emission across the field of view varied with about a ratio of seven between the brightest patch and darkest patch of the image, but in all cases the signal-to-background ratio was at least 2. Each patch was defined as seven pixels from the optical fiber bundle.

Fig. 7

(a) Pseudo-white-light image of Group 2 of the 1951 USAF Resolution Target. (b) Pseudo-white-light, false-colored image captured during navigation in an ex vivo porcine fallopian tube. (c) Monochrome reflectance image showing the cell collection wire (indicated by red arrow) extended during scrape biopsy in an ex vivo porcine fallopian tube. (d) Monochrome fluorescence image taken with 488-nm illumination and a 500-nm long-pass filter in an ex vivo porcine fallopian tube (Video 1, MP4, 14 MB [URL: https://doi.org/10.1117/1.JBO.26.7.076001.1).

Feasibility testing in ex vivo human fallopian tube tissue confirmed that the endoscope successfully fit through sections of the isthmus and ampulla for imaging. Reflectance and fluorescence images were acquired in ex vivo porcine fallopian tube tissue as shown in Fig. 7. This includes pseudo-white-light images captured within the tubal lumen during navigation [shown in Fig. 7(b)], monochrome reflectance images captured during cell collection [shown in Fig. 7(c)], and fluorescence images captured during navigation [shown in Fig. 7(d)]. These images were processed with a 1 pixel-width Gaussian blur function to de-emphasize the fiber bundle honeycomb appearance. Figure 7(d) is a frame from a video captured in fluorescence imaging tests, which demonstrates translation of the endoscope during navigation and visualization of fluorescent dye (Video 7D). We encountered some problems with intense light reflections and floating debris previously reported in other falloposcopy procedures.^35,39,42 However, these phenomena were managed with the introduction of air and saline into the introducing catheter and cell collection tubing, respectively, for distending the tube and clearing debris from the optics while imaging.

3.2.

Cell Collection

Statistical analysis of the number of cells collected with Nitinol wires of various designs revealed that there was no statistically significant difference in the average number of cells collected by wires of each tested shape ($P=0.92$, Table S1 in the Supplementary Material) diameter ($P=0.84$, Table S2 in the Supplementary Material), and surface finish ($P=0.87$, Table S3 in the Supplementary Material), nor in the number of cells collected from isthmus, ampulla, or infundibulum sections of ex vivo porcine fallopian tubes. In all cases, on the order of ${10}^5\text{cells}$ per tissue scrape could be obtained.

Figure 8 shows images of samples from the test biopsy procedures. As shown in Fig. 8(a), H&E-stained cross-sections of scraped fallopian tube show clean removal of the FTE with no histologically apparent damage to deeper tissues. Tissue clumps collected prior to dissociation and individual cells retrieved after dissociation are shown in Figs. 8(b) and 8(c), respectively. Figure 8(d) shows dissociated secretory and ciliated epithelial cells with nuclear material, PAX8, and acetylated tubulin labeled using standard immunofluorescence procedures. In addition, Fig. S1 in the Supplementary Material shows dissociated FTE cells with stained vimentin and cytokeratin-7, confirming that they are of epithelial origin. Multiple samples suggest that we are collecting mostly intact epithelial cells from the FTE in each biopsy procedure, sufficient in quantity and quality for multiple types of analyses.

Fig. 8

(a) H&E-stained cross-section of biopsied fallopian tube tissue showing the atraumatic removal of the epithelial layer of cells on the right side of the image. (b) H&E-stained clump of tissue prior to the dissociation procedure. (c) H&E-stained cells spread in monolayer on a slide after dissociation. (d) Image from immunofluorescence experiments on dissociated cells showing secretory cells with labeled PAX8 (magenta), ciliated cells labeled with acetylated tubulin (green), and all nuclei labeled with DAPI (blue). Arrows are used to indicate secretory cells (white arrow), primary cilia (yellow arrow) on secretory cells, and multi-cilia (orange arrow) on ciliated cells collected from the FTE.