1. Introduction

Tethered capsule endomicroscopy (TCE) is a noninvasive technology for imaging the upper gastrointestinal (GI) tract of unsedated patients, that uses a swallowable optomechanically-engineered tethered capsule [1–4]. After being swallowed, the capsule descends through the upper GI tract through a combination of peristaltic activity and gravity and can be retracted manually by the operator [1–3]. TCE’s sedation-free approach allows the procedure to take place in an outpatient setting and be conducted by non-physician clinical staff [5], decreasing procedure costs and patient discomfort in comparison to standard of care white light endoscopy (WLE). Optical coherence tomography (OCT) is the most commonly used microscopic imaging method in tethered capsules [2,3,5–15]. OCT TCE enables comprehensive, depth-resolved microstructural imaging of long segments of the upper GI tract within a few minutes, opening up the possibility of screening for multiple diseases of the esophagus [2,3,5,6,11,12,14,15], stomach, and small intestine [10].

While OCT is powerful as a microscopic, structural imaging technology, morphology alone does not completely capture all available diagnostic features in the GI tract. Other features such as genetic mutations [16], molecular receptor [17], cell surface glycoprotein [18], and enzymatic expression [19] provide key insights into cell types, states, and malignant potential that are helpful for more precise diagnosis/prognosis of many GI tract conditions. These features are potentially accessible by fluorescent labeling [18–21] or the use of fluorescent molecular beacons [22].

A growing number of exogenous fluorescent markers have been developed for applications in the GI tract, including peptides, lectins, antibodies, and activatable probes [21]. Peptides with differing fluorescent labels have been applied for early detection of esophageal adenocarcinoma (EAD) and colorectal cancer [23–25]. Lectins have been conjugated with fluorophores and reported to aid in detection of neoplasia in Barrett’s Esophagus (BE) [18,26], similar to antibody-based fluorescent markers [27]. Also, enzyme activatable fluorescent probes show promising potential for disease detection in the upper GI tract [28].

Merging OCT with fluorescence (FL) imaging in tethered capsules therefore can increase the diagnostic capacity of TCE by informing on disease progression and response to therapy. Multi-modality TCE enables molecular information to be placed in the context of microstructure, which is important when attempting to ascertain where the fluorescence emanates from anatomically. Combining OCT with FL also enables better quantification of fluorophore concentration as the FL intensity can be calibrated based on the distance from the probe to the tissue, as determined by OCT [29].

OCT has previously been combined with FL imaging, demonstrating the potential of this complementary multi-modal imaging technology on the bench-top [30,31]. To date, several endoscopic applications of combined OCT and FL imaging have been reported for intra-coronary imaging [32–34], GI tract imaging [35,36], and airway imaging [37,38], among others.

In this paper, we introduce a new OCT-FL TCE imaging system and device that enables complementary, co-localized, and simultaneous microstructural (OCT) and molecular (FL) imaging in a single tethered capsule. Performance of our OCT-FL TCE technology is validated in phantoms, and feasibility of OCT-FL TCE is demonstrated in a methylene blue (MB)-stained swine esophageal injury model, ex vivo and in vivo. With the future anticipated translation of targeted fluorescent agents, this new OCT-FL TCE technology may enable more precise evaluation of the GI tract for a variety of neoplastic and inflammatory diseases.

2. Methods

2.1 Imaging system and device

The dual-modality OCT-FL TCE imaging system consists of two directly interfacing sub-units (OCT imaging system and FL imaging system) and an attachable imaging probe (OCT-FL TCE device). Schematic diagrams of all three sub-components are depicted in Fig. 1. Figure 2 depicts photographs of the OCT and FL imaging systems (Fig. 2(a)) and the OCT-FL TCE device (Figs. 2(c) and (d)), as well as a detailed schematic of the OCT-FL TCE device (Fig. 2(b)).

 

Fig. 1. Schematic diagram of OCT-FL TCE imaging system. (a) OCT imaging system, (b) FL imaging system, (c) OCT-FL TCE device. TLS – tunable light source, BS – fiber-optic beam splitter (50/50), CIR – fiber-optic circulator, L – lens, TS – linear translation stage, M – mirror, PBS – polarizing fiber-optic beam splitter, BD – balanced photo detector, DAQ – data acquisition board, MPU – motor power unit, PC – personal computer, CPU – central processing unit, SMF – single-mode fiber, PMF – polarization-maintaining fiber, DCF – double-clad fiber, MMF – multi-mode fiber, EL – electrical connection, FL-LS – fluorescence light source, SHU – fiber-optic shutter, PM – power meter, WDM – wavelength division multiplexer, DCFC – double-clad fiber coupler, EF – emission filter, PMT – photomultiplier tube.

Download Full Size | PDF

 

Fig. 2. OCT-FL TCE imaging system and device. (a) Photo of the OCT-FL TCE imaging system comprising the OCT imaging system, the FL imaging system, and the OCT-FL TCE device. (b) Schematic diagram of capsule. (c) Photo of 2 m long OCT-FL TCE device composed of electrical and optical connections, tether, and capsule. (d) Magnified view of (c) showing capsule incorporating focusing optics (ball lens), side-directing mirror surface (reflective prism), and micro motor. Scale bars: 5 mm.

Download Full Size | PDF

The OCT imaging system (Fig. 1(a)) is similar to previously published imaging systems of our group that can be found elsewhere [10,12,15]. In short, the system is based on a conventional swept source (SS-) OCT setup featuring wavelengths spanning 1310 ± 50 nm (resulting in an axial resolution of ∼5.5 µm in tissue of a refractive index of n = 1.4); a 100 kHz A-scan rate tunable light source (AXP50124-8, Axsun Technologies, USA); a single-mode fiber (SMF) based Mach-Zehnder interferometer (SMF-28); an indium gallium arsenide (InGaAs) polarization diverse dual-balanced detection unit (to avoid image artifacts due to stress-induced polarization changes in the optical fiber of the TCE device [39]; FD80F8, Fermionics Opto Technology, USA); and high-speed data acquisition (DAQ) electronics (AXP50124-8, Axsun Technologies, USA).

The interferometer’s sample arm light is guided towards the FL imaging system (Fig. 1(b)), which features a fiber-coupled 647 nm narrowband light source (OBIS FP 647LX, Coherent, Inc., USA) for excitation of MB. The light source’s output passes a mechanical beam shutter (SH-200-635-4, OZ Optics Ltd., CA) that is part of the system’s clinical laser safety circuit (LSC). The LSC is an essential requirement for a clinical class 3B device (ANSI), such as the FL imaging system. In the default state, the shutter is closed, prohibiting any subject exposure. Emission is enabled when all of the following three requirements are simultaneously met: i) the integrated light source power monitor records a power level below the specified safety threshold (e.g. <20 mW); ii) the scanning safeguard registers rotational beam scanning within the OCT-FL TCE device (via probing the counter-electromotive force); iii) the device operator actively initializes emission by pressing a button on the FL imaging system. Light passing the open shutter is coupled back into an SMF and combined with the light from the OCT imaging system using a 2×1 wavelength division multiplexer (WD6513A, Thorlabs Inc., USA). Both the FL excitation and the OCT light are further guided through the single-mode port of a 2×2 double-clad fiber coupler (DCFC; DC1300LEFA, Thorlabs, Inc., CA). The single-mode emission port of the DCFC is where the fiber connection to the imaging device is established (Fig. 1). Light returning from the sample (including OCT, FL excitation and emission) is collected in the core (diameter = 9.0 µm; NA = 0.12) and the first cladding (diameter = 105.0 ± 5.0 µm; NA = 0.2) of the DCF and separated at the DCFC. The light in the core is guided back towards the OCT imaging system, where it is interfered with the reference arm light before being directed towards the OCT detection unit. Due to negligible spectral response of the InGaAs detection material below 800 nm, spectral filtering for the FL excitation and emission light is disregarded. The light in the first cladding is guided towards the FL detection unit, consisting of a fiber collimator, a ∼663-800 nm bandpass emission filter (FF01-731/137-25, Semrock Inc., USA) to filter the 647 nm FL excitation and super-1260 nm OCT light, and a photomultiplier tube (PMT; H14677-42, Hamamatsu Photonics K.K., JP). The spectral response of the silicon PMT further suppresses detection of OCT light in the FL detection unit. The analog PMT output is digitized at a 100 kHz sampling rate using a DAQ board (USB-6361, National Instruments Corp., USA) and the data is transferred to the PC in the OCT imaging system. The FL data is displayed in real time as a false color ring surrounding the co-registered grayscale OCT image data (Fig. 1(a)).

The OCT-FL TCE device is similar to previously reported TCE devices [10,12,15] and consists of a ∼2 m long, 1 mm diameter flexible tether terminated by an 11 × 25 mm rigid pill-shaped capsule (Figs. 1(c) and 2(b)-(d)). On the proximal end, the tether has two connectors that establish transfer of optical and electrical signals from the OCT and FL imaging systems to the OCT-FL TCE device (Fig. 1). The tether houses i) the same DCF as the DCFC in the FL imaging system and ii) electronic cables powering the micro motor assembly that performs the rotational beam scanning within the capsule [7]. The capsule incorporates the focusing optics as well as the beam scanning unit (Figs. 2(b) and (d)). An in-house fabricated 1.7 mm diameter ball lens on a 1.5 mm long glass spacer (excluding the ball lens diameter) is used to achieve the desired focal distances of ∼7.0 mm and ∼7.5 mm for FL excitation and OCT light, respectively (resulting in illumination/detection NAs of ∼0.05/0.1 for FL and ∼0.05/0.05 for OCT imaging). Thus, spot sizes of ∼25 µm and ∼30 µm (1/e²), can be achieved, respectively. The ball lens is created by melting a 1 mm diameter glass spacer, that has been spliced to the tip of the DCF, into a ball using a commercial fusion splicer (Fujikura Ltd., JP). The emitted light is then directed nearly perpendicular towards the cylindrical imaging window of the capsule (at an angle of 8° to avoid specular reflection from the surface) via a reflective prism. This prism is mounted to the shaft of a micro motor (Nidec Sankyo CMI Corp., JP) that is aligned co-axially to the beam path. Circumferential beam scanning is effectuated as the prism rotates. The prism spins at ∼39 Hz, which, given the 100 kHz sampling rate for OCT and FL detection, results in 2560 FL samples/OCT depth profiles per rotation (at a lateral sampling of ∼13.5 µm). The OCT-FL TCE device does not incorporate means for automated axial translation. Image acquisition along the third (axial) dimension is accomplished by manual operator pullback.

2.2 Data processing, display, and storage

The OCT-FL TCE imaging system provides real time data processing and display through a custom designed software that runs on the OCT imaging system PC and is operated from an integrated touch screen monitor. The online data preview displays the FL data in a false color ring (from low-to-high: black-to-blue-to-cyan-to-green-to-yellow-to-white) surrounding the co-registered grayscale OCT data (Fig. 3). The OCT data processing pipeline of the raw fringe data includes conventional SS-OCT pre-processing steps (background compensation, windowing, numerical dispersion compensation, Fourier transform) as well as polar-Cartesian coordinate scan conversion (Figs. 3(a) and (b)). The 1D FL data is directly mapped to the false color scale and scan converted as well (Figs. 3(a) and (b)). When data is actively being recorded, the raw OCT and FL data are continuously saved to a hard drive while the preview displays live image data.

 

Fig. 3. OCT-FL TCE data display of representative images from swine esophagus, in vivo. (a) Polar and (b) Cartesian representation of the same cross-sectional scan, depicting 2D grayscale OCT and 1D false color FL data. (c) 3D representation of the cross-sectional OCT map along the axial extension of the esophagus (inverted grayscale: low-to-high as black-to-white). (d) 3D representation of the FL surface map along the axial extension of the esophagus. Scale bars: 1 mm.

Download Full Size | PDF

2.3 Methylene blue fluorescence agent

The choice of MB as the initial fluorophore target for OCT-FL TCE was based upon several considerations. First, MB in its own right, is known to stain absorptive tissue such as metaplastic columnar epithelium, but not non-absorptive tissue such as squamous or dyspaltic epithelium [40,41]. Thus, it holds potential for diagnosis of BE and other inflammatory diseases (e.g. IBD). In the application of MB for chromoendoscopy [40–42], the positive selectivity for metaplastic epithelium was found to allow for differentiation from regions of reduced uptake corresponding to high grade dysplasia and malignancies [43]. Second, topical application of MB on the esophageal epithelium, as in chromoendoscopy, is reported to be safe, and thus clinically translatable [42]. Third, its FL excitation and emission maxima of ∼665 nm and ∼690 nm, respectively, match that of other targeted fluorescent probes for application in the upper GI tract [24,25,28].

Two performance tests were conducted on the lab bench to investigate i) the FL detection sensitivity for MB concentration in solutions and ii) the signal attenuation as a function of distance from the outer surface of the capsule imaging window.

First, MB powder (Sigma-Aldrich, USA) was mixed with deionized water to produce solutions of different concentrations in separate containers and the OCT-FL device was then sequentially submerged into the respective solutions. Data was acquired for each concentration by recording 100 consecutive OCT-FL frames containing 2560 FL samples while excluding a 120-sample-wide window that corresponds to the micro motor artefact.

Second, a semitransparent plastic capillary (2.7 mm outer, 0.3 mm wall thickness) was filled with a highly concentrated MB solution (1 mM, to compensate for high signal losses introduced by the capillary) and mounted parallel to the axial extension of the OCT-FL device in close proximity to the outer surface of the capsule imaging window. Next, the capillary was gradually moved away from the capsule with the help of a linear translation stage. Data was acquired at every step, and signal attenuation as a function of distance was determined by calculating means and standard deviations over a 200-sample-wide window over 10 consecutive frames.

2.4 Swine injury model

Yorkshire swine were imaged to test feasibility and performance of OCT-FL TCE towards clinical translation, an animal model with similar GI tract anatomy to humans. As mentioned above, MB is not taken up by non-absorptive squamous epithelium, but only by actively absorbing columnar epithelium [40,41]. Such, healthy swine esophageal tissue, which is entirely squamous, is not a good model to demonstrate performance of OCT-FL TCE in conjunction with MB staining. Due to the lack of availability of good swine models for human GI tract diseases [44], an alternative approach was pursued. We chose to injure the esophageal epithelium with cold biopsy forceps, thus creating superficial tissue trauma where MB was found to non-specifically adhere. This simple swine injury model using positive contrast MB staining sites allowed validation of our technology in human-like anatomy, ex vivo and in vivo.

All animal studies were performed according to the Massachusetts General Hospital’s (MGH) Institutional Animal Care and Use Committee under approved protocol 2019N000129.

2.5 Ex vivo esophagus study

For the ex vivo esophagus study, injury sites were induced along a significant length of the esophagus (∼10 cm, starting distally at the gastroesophageal junction (GEJ)). These injuries were achieved with a cold biopsy forceps through the working channel of a standard endoscope, while the swine was under general anesthesia. After sacrificing the swine, the esophagus was harvested, and the imaging study was performed immediately thereafter. The esophagus was mounted hanging vertically, along its axial extension, to mimic orientation of the human esophagus in an upright sitting or standing subject. The study protocol incorporated the following steps:

2.6 In vivo swine esophageal imaging study

A similar approach was used for the in vivo swine esophageal imaging study. Biopsy sites were created endoscopically while the swine was under general anesthesia. Then, the following protocol steps were carried out in the living animal:

3. Results

3.1 Methylene blue fluorescence performance evaluation

The results of the two FL imaging performance tests are summarized in the graphs in Fig. 4. Figure 4(a) depicts the PMT readout as a function of MB solution concentration. The PMT signal background was 1.31 V. The expected linear relationship between the FL signal and concentration can be observed. With MB’s low FL quantum yield between 0.03-0.06 [45], the limit of detection (LOD) was calculated to be ∼60 nM. Figure 4(b) depicts the results from the distance attenuation experiment. The PMT readout is plotted as a function of distance between the outer surface of the capsule imaging window and the capillary. The signal attenuation follows a second order polynomial with the half-intensity point at ∼3.5 mm of linear translation. A representative Cartesian OCT-FL image is depicted for each of the two experiments. For the concentration experiment, a 360° uniform blue color in the FL part of the image (excluding the 120-sample-wide micro motor artefact) indicates the homogenous dye distribution within the solution (Fig. 4(c)). For the attenuation experiment, the location of the capillary can not only clearly be observed in the FL part of the image (green/cyan area) but also in the OCT image data (Fig. 4(d)). Other structures visible in the OCT images correspond to the micro motor cables and the mounts fixing the capsule for the respective experiment.

 

Fig. 4. Results of the methylene blue (MB) fluorescence performance evaluation experiments. (a) PMT readout plotted as a function of MB solution concentration (error bars indicate ±1 standard deviation). (b) PMT readout plotted as a function of distance between the outer surface of the capsule imaging window and the capillary (error bars indicate ±1 standard deviation). (c) Representative Cartesian OCT-FL image from the concentration experiment. (d) Representative polar OCT-FL image from the attenuation experiment.

Download Full Size | PDF

3.2 Ex vivo esophagus study

Representative imaging results of the ex vivo esophagus study are depicted in Fig. 5. The top row of images (Figs. 5 (a)-(d)) presents en face views of a ∼6 cm long esophageal tissue section, featuring a direct comparison between the respective imaging modalities. A photograph of the cut open section in Fig. 5(a) provides the ground truth validation regarding the 9 biopsy induced injury sites. Figures 5(b) and (c) depict the FL surface map and the OCT en face projection map (depth-averaged over the entire imaging depth), while Fig. 5(d) depicts a composite image of Figs. 5(b) and (c) in a transparent overlay. Even though the injury sites are identifiable in the OCT en face projection as brighter areas, visibility is greatly enhanced in the FL surface map. All 9 sites (indicated by white arrow heads in Figs. 5(a)-(c)) can be matched directly by location and shape to the ground truth photograph. The middle row of images (Figs. 5(e)-(h)) presents magnified views of the region of interest (ROI) indicated by the white square in the images of the top row. This ROI highlights an injury site and provides a close-up of the differences between the respective imaging modalities. It is noted that the OCT en face projection of the ROI (Fig. 5(g)) is depth-averaged over the first 300 μm of tissue and thus appears different from the marked ROI in Fig. 5(c). The composite overlay (Fig. 5(h)) further illustrates that FL and OCT do not highlight the same tissue structures and thus provide complementary contrast. The bottom row of images (Figs. 5(i)-(k)) presents polar OCT cross-sections with the corresponding FL signal line on top. The three OCT cross-sections correspond to the respective locations indicated in Figs. 5(b) and (c) (i.e. dotted white lines). Injury sites identified in the cross-sections are indicated by black arrow heads, matching those indicated in Figs. 5(a)-(c). A biopsy trauma site typically presents with inhomogeneity in the epithelial layer and instances of emerging lamina propria in OCT cross-sections. The OCT signal intensity distribution is also less homogenous than in surrounding healthy tissue, and an overall increased intensity at the injury site causes reduced signal content from deeper tissue areas (i.e. shadowing). The identified locations of the injury sites in the OCT cross-sections match the areas of increased FL signal.

 

Fig. 5. Representative imaging results from the ex vivo esophagus study. (a) Photograph of a cut open esophageal tissue section for ground truth validation of the biopsy induced injury sites. (b) FL surface map and (c) OCT en face projection map (depth-averaged over entire imaging depth) of the same esophageal section. White arrow heads in (a)-(c) indicate matching injury sites in all three images. (d) Composite image overlaying (b) and (c), showing perfect co-registration. (e)-(h) Representative magnified views of the injury site indicated by the white square in (a)-(d) (OCT en face map here is depth-averaged over the first 300 μm of tissue from the epithelial surface). (i)-(k) Representative polar OCT cross-sections (including FL signal on top) of ROIs indicated in (b) and (c). Black arrow heads in (i)-(k) indicate matching injury sites between en face and cross-sectional images. Scale bars (a)-(d): 5 mm; (e)-(k): 1 mm.

Download Full Size | PDF

3.3 In vivo swine esophageal imaging study

Representative imaging results of the in vivo swine esophageal imaging study are depicted in Fig. 6. Figures 6(a)-(d) present en face views of a ∼10 cm long section of the esophagus, again featuring a direct comparison between the respective imaging modalities. A photograph of the excised and cut open section in Fig. 6(a) provides the ground truth validation regarding the 11 biopsy induced injury sites. Figures 6(b)-(d) depict FL surface maps and the OCT en face projection map (depth-averaged over the entire imaging depth). The first surface map (Fig. 6(b)), showing a homogenous dark blue color, was acquired during the background pullback without presence of MB. Similar to Figs. 5(b) and (c), the injury sites are already identifiable in the OCT en face projection at detailed inspection; however, visibility is greatly enhanced in the FL surface map. All 11 sites (indicated by white arrow heads in Figs. 6(a), (c), and (d)) can be matched directly by location and shape to the ground truth photograph. Again, MB uptake/accumulation is mainly observed in the border areas of the injury sites, as highlighted in the representative magnified view of an injury site in Fig. 6(h), an ROI indicated by a white square in Fig. 6(c). Also, unwashed excess dye that was trapped in tissue folds, causing bright linear structures along the pullback direction, is observed at two instances in Fig. 6(c) (indicated by white asterisks). Figures 6(e) and (f) present recorded video endoscopy footage at the same esophageal location before and after the MB staining procedure. Here, the visible injury site shows blue surface staining in Fig. 6(f). Figure 6(g) also depicts video endoscopy footage of the TCE procedure at an MB-stained injury site. Figures 6(h) and (i) show magnified views of the ROIs indicated by the white square in Figs. 6(c) and (d). Since the OCT en face projection of the ROI (Fig. 6(i)) is depth-averaged over the first 300 μm of tissue from the epithelial surface, it shows a difference in appearance to that of the marked ROI in Fig. 6(d). Also, both figures were postprocessed to crop a short axial segment of a non-uniform pullback artefact. Figure 6(j) depicts a composite image of Figs. 6(h) and (i) in a transparent overlay, again highlighting the complementary image contrast. Figures 6(k)-(m) present polar OCT cross-sections with the corresponding FL signal line on top. The three OCT cross-sections correspond to the respective locations indicated in Figs. 6(c) and (d) (i.e. dotted white lines). Injury sites identified in the cross-sections are indicated by black arrow heads, matching those indicated in Figs. 6(a), (c) and (d). Identical to the ex vivo esophagus study, biopsy trauma sites present with epithelial inhomogeneities, increased superficial signal, and shadowing in deeper tissue layers in the OCT cross-sections. The identified locations of the injury sites in the OCT cross-sections match the areas of increased FL signal. Figure 6(m) furthermore visualizes, in both the FL and the OCT images, the locations of the two tissue folds that trapped or contained more MB (indicated by black asterisks).

 

Fig. 6. Representative imaging results from the in vivo swine esophageal imaging study. (a) Photograph of a cut open esophageal tissue section for ground truth validation of the biopsy induced injury sites. (b) FL background surface map, (c) FL MB surface map, and (d) OCT en face projection map (depth-averaged over entire imaging depth) of the same esophageal section. White arrow heads in (a), (c), and (d) indicate matching injury sites in all three images. (e), (f) Video endoscopy footage showing an injury site before and after MB staining procedure. (g) Video endoscopy footage showing TCE procedure at an MB-stained injury site. (h), (i) Representative magnified views of the injury site indicated by the white square in (c) and (d) (OCT en face map here is depth-averaged over the first 300 μm of tissue from the epithelial surface). This same injury site is also indicated by a white square in (e) and (f). (j) Composite image overlaying (h) and (i). (k)-(m) Representative polar OCT cross-sections (including FL signal on top) of ROIs indicated in (c) and (d). Black arrow heads in (k)-(m) indicate matching injury sites between en face and cross-sectional images. Scale bars (a)-(d): 5 mm; (h)-(m): 1 mm.

Download Full Size | PDF

4. Discussion

We present OCT-FL TCE, a dual-modality approach for upper GI tract imaging that combines simultaneously acquired and co-registered, microstructural OCT and molecular FL imaging capabilities. We showcase a clinically viable, compact, portable imaging system and tethered capsule device and validate this technology in phantoms and in a swine esophageal injury model, ex vivo and in vivo. We demonstrate system performance and feasibility through application of MB as our fluorescent agent.

OCT-FL TCE with MB has potential for clinical translation in BE. Topical MB chromoendoscopy [40–42] has previously been shown to distinguish between nondysplastic/low grade BE and high grade dysplastic BE and cancer, based on differences in local MB uptake [43]. Since dysplastic BE can sometimes be difficult to appreciate in full OCT scans of the esophagus [46], the addition of a fluorescent marker to aid in interpretation could be highly valuable clinically. It should be noted that clinical translation of the presented technology will require an alternative approach for applying NAC and MB, as a spray catheter currently requires sedated endoscopy. Also, the OCT-FL TCE clinical procedure time is expected to be slightly greater than that of conventional OCT TCE, due to the time required to administer NAC and MB.

The potential applications of OCT-FL TCE extend far beyond esophageal MB. First, TCE has been shown to be capable of imaging other upper GI tract organs, including the stomach and small intestine [10]. Second, the presented modular imaging system design allows simple adaptation to any specific excitation or emission band within a wavelength range of 400-900 nm. Only the FL light source and the emission filter would need replacement, while the rest of the imaging system would remain unchanged, requiring minimal engineering effort. Hence, ongoing development efforts in the field of molecular targeted fluorescence labeling might allow for exploitation of our platform for a host of different applications and targeted molecular agents [21].

The complementarity of structural and molecular imaging is of critical importance for enhanced disease diagnosis in the GI tract. For example, when imaging near the GEJ, it is imperative to understand whether the signal arises from the distal esophagus or gastric cardia (most proximal portion of the stomach), as the prognostic significance of identifying intestinal metaplasia is different in these two organs [47]. Should the capsule only be able to detect fluorescence, the anatomical location at this transition zone could not be ascertained. Similar advantages for combining microstructure (OCT) and molecular (FL) information exist when detecting subsurface pathology that can be seen in neoplastic and inflammatory processes (e.g. Crohn’s disease).

In contrast to our group’s previous combined near-infrared FL and OCT investigations for intravascular coronary imaging [29], here, we have not integrated an intensity distance calibration for FL signal adjustment, which would incorporate luminal surface segmentation of the OCT image data. We found that the outer surface of the capsule experienced a very high degree of direct tissue contact. Hence, the FL signal was obtained from the same depth location, rendering a potential intensity distance calibration unnecessary. It should be noted that the esophagus probably presents a unique exception in comparison to other segments of the GI tract (stomach, small intestine, colon) concerning the degree of direct device/tissue contact. Future studies in the upper GI tract beyond the esophagus may necessitate FL intensity distance calibration [48].

One major advantage of OCT is its capacity to enable whole organ microscopic imaging [2,3,10,49], mitigating sampling errors that are inherent in standard endoscopic biopsy. The challenge with whole organ imaging is that the amount of data can be overwhelming, especially when attempting to render a diagnosis in near real time to effectuate a clinical decision. In the GI tract, accurate diagnoses are made by a limited number of expert OCT readers, hindering a broader clinical adoption of whole organ OCT in the GI tract. Thus, another potential role of FL in OCT TCE is to highlight diseased ROIs, decreasing the amount of image data that needs to be viewed and potentially lowering the complexity of image interpretation.

5. Conclusions

In this paper, we introduce dual-mode TCE through a combination of structural OCT imaging and tissue type targeted FL imaging. We present a modular, easily portable, plug-and-play system design of a clinically viable device. We validated our technology in phantom experiments and demonstrated feasibility in preclinical swine studies, ex vivo and in vivo. For this proof-of-principle investigation, we selected MB as our fluorophore of choice, as this approach has potential for direct clinical translation in BE. With the OCT-FL TCE technology, the repertoire of conventional OCT-based TCE is expanded beyond purely morphology to include additional molecular imaging, hence opening up the possibility for a superior diagnostic approach that leverages the complementarity of both imaging techniques. In combination with recently introduced and newly developed targeted molecular agents, OCT-FL TCE is likely to significantly improve our diagnostic capabilities for a wide array of GI tract diseases.