3.1.

Operating Room Environment

Falloposcopes are submitted to the sterilization department the day before the surgery. On the day of the surgery, the falloposcope is delivered to the OR with all other medical instruments necessary for the gynecological procedure. While the surgical technicians prepare for about 30 min, the research team sets up the falloposcope medical cart. When the patient is transported into the room by hospital bed, they are transferred onto the operating table and then anesthetized. Our team is asked to vacate the OR during that time to give the patient privacy. Once the patient is fully anesthetized, we are allowed to continue setting up the medical cart. There are a few minutes until the surgeon begins the time out, during which the OR team reviews the patient’s identity, the procedure, and the surgical site before the start of the procedure. At the surgeon’s discretion, after cervical dilation and prior to salpingectomy, the pilot study is performed. After the pilot study procedure’s 15-min time is up, the surgeon continues with the procedure, and the research team waits to get explanted tissue and retrieve the pilot study instruments and tools that were used.

We quickly learned that the varying sizes and configurations of the ORs, as well as surgical staff preference, meant that we needed to be flexible to accommodate any placement within the OR that was asked of the research team, requiring such minor but critical items as a long extension cord. Communications with staff, including clear directions from the surgeon, are also critical. For example, in first pilot study, although the research team was out of the OR, both the primary and secondary falloposcope cases were opened and placed on tables with other medical tools to be used in surgery. We were able to share a warning about the delicacy of the falloposcopes with one surgical technician, but as the procedure was about to start, more surgical technicians came in and the one who performed the preparations left. Communication at that point was difficult because all unnecessary chatter must end once time out commences. We anxiously watched our microendoscopes be handled roughly, be placed in precarious positions, and have other medical tools piled on top of them after the procedure. For the next procedure, we requested one table to be solely dedicated to the research team’s equipment. We asked that all of the components be carefully removed from the case and placed onto the table, and the case itself handed back to the research team to avoid inadvertent damage to the microendoscope. Figure 2 shows the arrangement of the falloposcope and accessory items (introducing catheter and hysteroscope) in the sterilization cases.

Fig. 2

Left: sterilization case with disposable instrumentation, including the (1) falloposcope, (2) introducing catheter, and (3) introducing catheter components. Right: sterilization case with reusable instrumentation, including (4) hysteroscope camera, (5) light guide cable, (6) hysteroscope sheath, (7) hysteroscope telescope, and (8) additional necessary components.

A second challenge was providing a clear view of falloposcope navigation using the forward facing MFI system with only the 488 laser source on. At the medical staff’s request, we did not connect our falloposcope instrument cart to the OR’s room monitors. Instead, a second opposite facing monitor was added to the cart to allow the surgeon to visualize MFI navigation and OCT scans. After the first subject was completed, the surgeon stated that the imaging system frames per second (fps) was not fast enough (about 7 fps) for them to comfortably navigate the microendoscope. The original software had a frame rate inversely proportional to the size of the display window, so either the image could be large enough or the frame rate fast enough to be comfortable to the physician, but not both. To increase the fps, we reduced the temperatures in the cart by 10°C to increase performance of the graphical processing unit and central processing unit while selecting a reduced window size to maintain a stable 20 fps live feed. To increase the video size in a large, comfortably viewed window, we used alternative software to magnify and stream the camera output to a second monitor. Display settings were optimized in the software and the monitor to enhance contrast. During the debrief after implementing this solution, the surgeon said that they were comfortable with the frame rate and were able to successfully navigate the microendoscope.

After a few patients, we realized that it was beneficial to record footage of the hysteroscope camera feed, which is not mandatory and even when obtained could be difficult to access after the procedure. We added our own hysteroscope camera module, which had two video outputs: a primary output connected to an external capture card to record the procedures and a secondary output plugged into a proprietary digital to analog converter to allow the display to be fed into the OR monitors. The configuration of the hysteroscopic view on the OR monitors and the falloposcopic view on the large instrument cart monitor were acceptable to the surgeon. Figure 3 shows the arrangement of the instrument cart in the OR.

Fig. 3

Attending physician performing a hysterectomy; the research team observes after our first successful 15-min pilot study procedure was completed. Sources and detectors are located in the black medical rack.

3.2.

Hysteroscope Accessories

Our first few pilot studies had SSG introducing catheters combined with commercial off-the-shelf single-use medical device components from Qosina (Ronkonkoma, Long Island, New York). We assembled these components consisting of a hemostasis valve $y$ connector (Qosina 80325), Luer lock tubing connector (Qosina 71637), and Tuohy Borst valve adapter (Qosina, 80402), as shown in Fig. 4 (top), to enable saline flushing through the introducing catheter around the falloposcope inside. This assembly was determined to be suboptimal. After some falloposcope damage was noted, it was determined that interior stepped edges of the Tuohy Borst valve adapter and hemostasis valve $y$ connector were catch points for the endoscope’s distal face. The hemostasis valve $y$ connector was essential for the saline flush and prevention of backflow, but the Tuohy Borst valve adapter was not because its purpose was to lock the falloposcope into position, a functionality that the surgeon determined was not necessary. Moreover, the size and weight of the Qosina components that comprised the original introducing catheter assembly increased the difficulty of concurrently manipulating the hysteroscope, introducing catheter assembly, and falloposcope during the pilot study procedure. To complete the pilot study procedure, the surgeon had to enlist the aid of at least one other sterile person to manage use of all three independent instruments. The SSG catheter’s tip was opaque and difficulties in docking the tip to the ostium and having position awareness of the falloposcope’s tip were frustrating to the surgeon and surgeon’s aid. Our new configuration incorporates a modified Novy cornual cannulation set (Cook Medical J-NCS-504070). The Novy set already includes a $y$ connector and has a clear tip, giving the surgeon better visual awareness of the tool’s position. A short female Luer lock to male Luer lock connector (Qosina 17656) and two longer female Luer lock to male Luer lock connectors (Qosina 80379) (Fig. 4, bottom) were added to lengthen the catheter assembly, enabling the Tuohy Borst to tighten abound the microendoscope’s semirigid tubing. An added advantage was the Novy set’s 0.035” diameter guide wire, which could be used to check patency in the FT prior to falloposcope introduction.

Fig. 4

Introducing catheter components. Top: original configuration of a Tuohy Borst (Qosina 80402); Male Luer Lock Tubing Connector (Qosina 71637); and hemostasis valve $y$-connector (Qosina 80325), which connects to the 5.5 fr SSG catheter (Cook J-SSG-554086). Bottom: current configuration of a Tuohy Borst Adapter from the modified Novy cornual cannulation set, a female Luer lock to male Luer lock connector (Qosina 17656), two female Luer locks to male Luer lock connectors (Qosina 80379) connect to the 5.5 fr modified Novy cornual cannulation catheter (Cook J-NCS-504070) with included $y$-connector, inner catheter, and 0.35” guide wire.

3.3.

Endoscope Robustness/Fiber Breaks

In some procedures, we began with a fully functioning microendoscope, and by the time we returned to the lab to run ex vivo measurements, the microendoscope had lost functionality of one of the three major systems: OCT, illumination, or imaging. The two most common failures included breaks at the fiber splice junctions between the handle and instrument cart or at the rigid-to-flexible transitions of support material. In case of a structural failure of the primary falloposcope during sterilization, pre-operation setup, or the pilot study procedure, a back-up falloposcope was also sterilized and put on standby. However, the short time allotted for the pilot procedure (15 min) meant that switching out during the procedure was impractical. Stakeholders requested an increase in the robustness of the design. Because sterilization and human subjects approval were dependent on the exterior materials and sealing methods, major modifications in the falloposcope fabrication technique were limited as new, “from scratch” prototypes would have to be retested and reapproved.

In designing a more robust system, we can utilize commonly used fiber optic cable design practices based on telecommunication industry trends. Most field technicians that use fusion splicers in the field must use care and relieve potential future strain at the weak point by adding a splice protector with a junction box that holds the fiber’s strength member and fiber position in place to minimize tensile forces on the glass fiber. Our initial design tried to implement the fiber optic cable without strength members or a splice protector due to limited space and to minimize cost.

The proximal length of the microendoscope consists of a vinyl tube that contains the illumination fiber, imaging fiber bundle, and OCT fiber. It is permanently glued to the handle and is secured to the medical instrument cart during use. Unfortunately, the vinyl has a high friction coefficient and can cause a fiber splice to break if the fibers are accidentally pulled by the tubing itself when bent or unbent. Also, during sterilization and pre-operation setup, the end of the tubing is unattached, so if the fibers are pulled, the tensile forces propagate through the glass fiber core/cladding, a mechanical weakness of glass. Our mitigation strategy, as shown in Fig. 5, was adding additional smaller tubes of vinyl around the fibers where they proximally exit the vinyl tubing. These tubes act as a tight buffer that limits bend radii, eases stress on the splice points, and makes for convenient handling and storage of the fibers. Another increase in robustness was with the addition of splice junction reinforcements, as shown in Fig. 6, which consisted of a snug fitting polyimide tubing (Microlumen, 110-I) and adhesive (Loctite^® 4013 cyanoacrylate liquid adhesive) and prepping the surface by cleaning the substrate and adherent with acetone and then prepping the surfaces with primer meant for adhering low-energy plastics (Loctite^® 7701™ Prism^® medical device adhesive primer), which increased the tensile strength of the bond by $\sim 150\%$ compared with a fusion splice alone.

Fig. 5

Close up of the protected proximal fibers bridging the medical rack and the endoscope’s vinyl tubing.

Fig. 6

Standard splice protector has a long rigid rod creating a rigid to flexible transition that may become a breaking point at the splice junction. The custom splice protector is semi-rigid and shorter and has eliminated breakage at the splice junction.

In previous distal tip/handle designs, the PEEK MLE was glued inside a 304 stainless steel (SS) hypotube of 0.0730″ diameter and 14 cm length extending from the handle. The purpose of the hypotube was to strengthen the transition from the handle to the thin flexible insertable portion of the falloposcope and to provide a larger diameter and crush resistant section for locking the Tuohy Borst valve. However, the rigid hypotube in turn created a failure point at the transition from its rigid end to the flexible falloposcope. Our modification, as shown in Fig. 7, was to replace the hypotube with two layers of medical grade tubing. The larger tube of shorter length was a custom construction of polyether block amide (PEBA) elastomer, an SS braid, and a polytetrafluoroethylene (PTFE) liner. The smaller tube of longer length was constructed of polyimide, an SS braid, and a PTFE liner (Microlumen, 390-VII). The new design is crush resistant yet flexible enough to provide strain relief to the working length of the microendoscope. Another rigid to flexible junction is at the very distal tip where the MLE meets the SS ferrule that is used for bonding the fibers and pull wires in place. Operational adjustments, including the new clear-tipped introducing catheter, which helps avoid forcing the falloposcope tip against uterine tissue, have eliminated failures at this location. However, a more robust solution may be to eliminate the ferrule by bonding the fibers, distal tip optics, and pull wires directly to the PEEK MLE material. Glass-PEEK bonding can be accomplished with Loctite^® 4013 cyanoacrylate liquid adhesive;²³ however, SS pull wires would likely require a non-adhesive method of distal fixation.

Fig. 7

Top: the falloposcope’s proximal working length was initially designed with a SS hypotube to provide a surface for catheter components to slide onto. This created a rigid to flexible transition that was a kink point. Bottom: the new version has a two-layered approach to go from a more rigid braided PEBA elastomer to a semi-rigid braided polyimide section serving the same function while also acting as a crush-resistant strain relief for the fiber.

3.4.

Optimization

The falloposcope software originally consisted of a combination of laboratory-written code for OCT (Visual Studio C++, Microsoft, Redmond, Washington, United States) and vendor software for the MFI imaging (Ocular, QImaging, Surrey, BC Canada), which performed well in the laboratory. A detailed standard operating procedure was written for the falloposcope instrument cart operator for obtaining images and communicating with the surgeon. The time restrictions and stress of the OR environment brought out the limitations of both the laboratory-written software and the vendor software under unexpected conditions. To fix bugs, errors and vulnerabilities of our laboratory-written software, we encouraged research personnel to provide code reviews and feedback. We improved the program’s quality, reliability, and functionality by

All of these changes to the falloposcope software both expedited and standardized the image-collection process, which is crucial in the OR.

3.5.

Fiber Bundle

The fiber bundle is made of 3000 fibers that are melted together and are coherent in that each core on one end carries its signal to the other end with minimal crosstalk and in the same relative spatial position. It is used as a relay system to carry the image created by the distal GRIN lens to a medical instrument cart, where a microscope objective and camera lens magnify the image onto the camera sensor. Unfortunately, the individual fibers are not uniform in loss, and variances in intensity occur across the entire bundle, as well as between each falloposcope. To mitigate this variation, flat field images were taken by securing an opal diffusing glass at the falloposcope’s working distance (WD) 5 mm that was back illuminated by a 150 watt quartz halogen source (Dolan-Jenner, Fiber-Lite Mi-150) with single gooseneck fiber optic cable (Dolan-Jenner, BGG1818) to create a near Lambertian source. Test images were obtained of a Ronchi ruling on white opal diffusing glass in the same configuration. Dark images were taken under zero illumination conditions. The test images, $R$, are corrected by a flat field correction by

Fig. 8

(a) The raw image is of an opal diffusing glass backed Ronchi ruling. (b) The flat field image of the fiber bundle shows that the relative illumination from core to core is not uniform. Using a flat field corrected test image (c), panels (a)–(c) are normalized for easier viewing.

3.6.

GRIN Lens and Stray Light

Other improvements were made to reduce the background noise, thus improving the signal-to-noise ratio of weak autofluorescence. The first source of noise was caused by stray illumination light coupling into the imaging GRIN lens. The initial prototype illumination fiber (Polymicro FSU 80/85/100 core/cladding/coating) was selected for its small OD, high 0.66 numerical aperture (NA), and high hydroxyl composition enabling efficient propagation of visible laser illumination through the working length of the microendoscope. This fiber was spliced to a 50/125/250/900 core/cladding/coating/buffer OM3 GRIN MMF, which was connected to the output fiber from the beam combiner in the instrument cart. Unfortunately, the polymicro fiber became obsolete before the beginning of the pilot study. Its selected replacement was a custom bare fiber (Fiber Technology Incorporated, 85/100 core/cladding bare borosilicate fiber), which retained the small diameter necessary to traverse the MLE. However, because the custom bare fiber did not have a coating, it was more likely to suffer from fine cracks along the surface due to micro bending, compression, and tension as the fiber was handled during installation. The telltale sign of these fine cracks was illumination light leakage out along almost the entire working length, as shown in Fig. 9. The leaked light transmits directly into the surrounding PEEK, causing autofluorescence of the material. Some of the PEEK autofluorescence was collected by the GRIN lens and propagated through the fiber bundle, causing a high background signal in the images. Our solution to this stray light problem was to use blackened PET heat shrink to hold the GRIN lens to the imaging fiber bundle, blocking stray light collection. Future iterations of the falloposcope will utilize black-colored polymer, assuring that any stray light and background autofluorescence is absorbed.

Fig. 9

(a) Light leaks out of the thin cladding bare fiber and into fluorescent PEEK MLE. (b) Our solution was to replace the clear heat shrink covering the OD of the GRIN lens with black heat shrink.

3.7.

Proximal Imaging System Optimization

Additional background noise was discovered in the medical instrument cart’s proximal imaging system. This system begins at the SMA905 port where the microendoscope’s imaging bundle is connected. The bundle’s cores are imaged onto a sensor after passing through a filter wheel. The open cage system allowed bright illumination in the OR to couple into the imaging system. The unwanted background was resolved by closing the cage system with black panels and threaded barrels when possible.

Further optimization was made to the proximal imaging system. The benchtop prototype system utilized a ultra-violet (UV) achromat (OptoSigma, NUDL-30-150P) as the camera focusing lens because this system included a 275 nm laser source. It was necessary to change the position of the objective when images were taken with various light sources between 275 and 638 nm, as well as when a filter was introduced to the optical path for fluorescence images. In the updated clinically ready system, the sources selected were 405, 488, 520, and 642 nm, with UV avoided due to safety concerns of applying this light to reproductive tissue. The UV achromat remained in the system because the aberrations caused by it were not significant except for chromatic focal shift, which could be compensated by changing the objective position. The limited 15 min clinical time, however, made changing the objective position for every measurement impractical. The UV achromat was replaced with a better performing visible (VIS) achromat (Thorlabs, AC254-150-A), and a broadband (${\mathrm{MgF}}_2$) coated blank (Edmund Optics, No. 32-951) was installed into the filter wheel for reflectance measurements. These changes allowed us to collect reflectance and fluorescence measurements without having to change the objective’s position. The proximal imaging system optical design is shown in Fig. 10. Finally, the entire proximal imaging system was secured onto a small platform with shock absorbing Sorbothane feet (Thorlabs AV3) to ruggedize for transport to the OR.

Fig. 10

Simplified representation of the proximal imaging system in the medical cart to show how light from the imaging fiber bundle is collimated, filtered, and then focused onto an image sensor.

3.8.

Illumination System

The goal of our illumination system is to have a large angle, top-hat beam profile from the perspective of the imaging system, so the object plane has nominal illumination uniformity. The small size of the endoscope permits only one illumination fiber, leading to asymmetric illumination and making a large angle illumination more critical at the short WDs employed. The illumination fiber’s 0.66 NA provides an $\sim 83\mathrm{deg}$ illumination angle, sufficient to cover the field of view of the image system in the minimum WD of 5 mm in air or 6.7 mm in 0.9% saline solution. The illumination laser sources are coupled into SMFs with small NAs. Once the SMF is coupled to the falloposcope’s illumination fiber, mode mixing occurs and the output beam profile broadens. However, the length of illumination fiber, 1 m, is not long enough to provide complete mode mixing and a flat-top beam output. The first falloposcope prototype used a miniature mode scrambler to force microbends in the illumination fiber to promote rapid mode-mixing. This system was effective and had the added benefit of permitting a varying amount of mode scrambling and consequently output beam NA. However, the microbends significantly reduced the longevity of the fiber, resulting in failures. The second fiber selected was connected to the OM3 by the manual splice method, which in turn meant that mode-scrambling was no longer necessary as the accidental axial position mismatch that occurred during manual splicing led to rapid, albeit lossy, mode-mixing. When the new third illumination fiber was selected, we wanted to avoid utilizing the mode scrambler, so it was decided to exchange the OM3 GRIN MMF for an SMF-28 fiber for better fiber-coupling. The beam profile of the SMF-28 is Gaussian, and splicing it to the 0.66 NA borosilicate fiber creates a broad and smooth, albeit non-uniform, Gaussian profile on the tissue surface. To assure comparable performance of all falloposcopes, the current of each laser sources is adjusted to normalize the output power, controlling for varying losses from endoscope to endoscope.

3.9.

Initial Performance with Improved Prototype

The improvements made to the falloposcope system eliminated points of failure and enhanced operational usability. The modifications to the system successfully allowed the full 15 min of allowable OR time to be utilized, thus increasing the quantity of data acquired. To be classified as non-significant risk, this study was performed on volunteers with no evidence of, or elevated risk for, ovarian cancer. Therefore, we cannot yet evaluate the falloposcope’s diagnostic potential. However, preliminary in vivo imaging capabilities are shown in Fig. 11. A reflectance image with 488 nm illumination, extracted from an in vivo video recording, is shown in Fig. 11(a) revealing the FT lumen as well as a protrusion in the wall of the FT. An OCT image of the proximal FT is shown on the right, revealing overlapping FT plicae.

Fig. 11

(a) Example reflectance image obtained with 488 nm laser illumination. The lumen (arrow) as well as a protrusion in the wall (arrowhead) are shown. The high reflectance streak in the center is specular reflectance. Gaussian blur is applied to reduce the honeycomb artifact. (b) Example OCT image, with overlapping FT plicae (arrows show examples). Histogram adjusted to increase contrast.

Ex vivo imaging has been performed with the falloposcope and some initial images published.²¹ The resolution of the forward looking imaging system is severely limited by the 3000 pixels of the miniature imaging fiber bundle; thus these images are a much lower visual quality than those taken previously with a tabletop MFI system.¹⁴ However, ex vivo and in vivo multispectral fluorescence and reflectance images taken with the falloposcope are comparable.

Falloposcope OCT images are similar in imaging depth and axial resolution to ex vivo images obtained with commercially available OCT tabletop systems. However, falloposcope images are lower lateral resolution due to the small diameter of the falloposcope and a conservative choice of NA to assure a long depth of focus. Most importantly, the falloposcope has no distal scanning mechanism, and 2D images are obtained by manual pull back. This method can cause repeated or discontinuous a-scans due to non-uniform pull back speed, a situation that is exacerbated in in vivo versus ex vivo imaging. In the future, we intend to implement an automated pull back system to provide a constant speed through acquisition.

We have described the iterative prototyping process utilized to transition the engineering design work from the laboratory to the clinical environment. The changes made to the microendoscope system, driven by stakeholder input, resulted in significant operational success. Notably, we have not experienced any broken fibers in the falloposcope during sterilization, pre-procedure operations, or normal in vivo use after implementing the design changes stated in this manuscript. We have plans for future incremental upgrades to the overall system, including designing a smaller profile and lighter weight handle, improving the fiber protection system, and developing a variant of the falloposcope that collects cells for cytology.