The assembly and use of a multimodal microendoscope is described which can co-register superficial tissue image data with tissue physiological parameters including hemoglobin concentration, melanin concentration, and oxygen saturation. This technique can be useful for evaluating tissue structure and perfusion, and can be optimized for individual needs of the investigator.
Recent fiber-bundle microendoscopy techniques enable non-invasive analysis of in vivo tissue using either imaging techniques or a combination of spectroscopy techniques. Combining imaging and spectroscopy techniques into a single optical probe may provide a more complete analysis of tissue health. In this article, two dissimilar modalities are combined, high-resolution fluorescence microendoscopy imaging and diffuse reflectance spectroscopy, into a single optical probe. High-resolution fluorescence microendoscopy imaging is a technique used to visualize apical tissue micro-architecture and, although mostly a qualitative technique, has demonstrated effective real-time differentiation between neoplastic and non-neoplastic tissue. Diffuse reflectance spectroscopy is a technique which can extract tissue physiological parameters including local hemoglobin concentration, melanin concentration, and oxygen saturation. This article describes the specifications required to construct the fiber-optic probe, how to build the instrumentation, and then demonstrates the technique on in vivo human skin. This work revealed that tissue micro-architecture, specifically apical skin keratinocytes, can be co-registered with its associated physiological parameters. The instrumentation and fiber-bundle probe presented here can be optimized as either a handheld or endoscopically-compatible device for use in a variety of organ systems. Additional clinical research is needed to test the viability of this technique for different epithelial disease states.
Fiber-bundle microendoscopy techniques typically analyze in vivo tissue using either imaging techniques or a combination of spectroscopy techniques.1-3 One such imaging technique, high-resolution fluorescence microendoscopy, can image apical tissue micro-architecture with sub-cellular resolution in a small, microscale field-of-view, using a topical contrast agent such as proflavine, fluorescein, or pyranine ink.1,3-11 This imaging modality has shown promising clinical performance in qualitatively differentiating diseased and healthy epithelial tissue in real-time with low inter-observer variability.8 Occasionally, investigators will use high-resolution fluorescence microscopy data to extract quantitative features such as cell and nuclear size or gland area, but this remains a primarily qualitative technique targeted towards visualizing tissue morphology.1,3,8-10 On the other hand, spectroscopy techniques, such as diffuse reflectance spectroscopy, are targeted towards providing functional tissue information and have shown promising clinical performance in quantitatively identifying cancerous lesions in multiple organs.2,12-15
Therefore, there is a need for a device incorporating both types of modalities to potentially further reduce inter-observer variability, maintain real-time visualization of tissue micro-architecture, and provide a more complete analysis of tissue health. To accomplish this goal, a multimodal probe-based instrument was constructed that combines two modalities in a single fiber-optic probe: high-resolution fluorescence microendoscopy and sub-diffuse reflectance spectroscopy.11 This method co-registers qualitative high-resolution images of apical tissue morphology (structural properties) with quantitative spectral information (functional properties) from two distinct tissue depths including local hemoglobin concentration ([Hb]), melanin concentration ([Mel]), and oxygen saturation (SaO2).11,12,16 This specific sub-diffuse reflectance spectroscopy modality uses two source-detector separations (SDSs) to sample two unique tissue depths to provide a more comprehensive picture of tissue health by sampling down to the basement membrane and underlying tissue stroma.11
The fiber-probe consists of a central 1 mm-diameter image fiber with approximately 50,000 4.5 µm diameter fiber elements, a cladding diameter of 1.1 mm and an overall coating diameter of 1.2 mm. The image fiber is surrounded by five 200 µm diameter fibers with cladding diameters of 220 µm. Each 200 µm multimode fiber is located a center-to-center distance of 864 µm away from the center of the image fiber. Each of the 200 µm multimode fibers are 25° apart. Using the leftmost 200 µm multimode fiber as the "source" fiber, and the additional three 200 µm multimode fibers as the "collection" fibers, this geometry necessarily creates three center-to-center SDSs of 374 µm, 730 µm, 1,051 µm, and 1,323 µm. The fiber tips are enclosed in a cylindrical metal casing that keeps the distances between fibers constant. The diameter of the cylindrical metal casing is 3 mm. The distal end (towards the fiber-optic probe tip) of the fiber-optic probe is 2 feet long. The probe then separates into the six respective individual fibers at the proximal end (towards the instrumentation) which is an additional 2 feet long, for a total length of 4 feet. Figure 1 shows a representation of the fiber-optic probe.
Figure 1: Fiber-optic probe design. The fiber-optic probe consists of one 1 mm-diameter image fiber and four 200 µm multimode fibers. This figure shows representations of (a) the metal end cap which constrains the geometry of the fibers at the probe tip to yield SDSs of 374, 730, and 1,051 µm with respect to the leftmost 200 µm multimode fiber (Scale bar ≈ 1 mm), (b) the fibers being constrained within the metal cap, showing the fiber cores, fiber cladding, and fiber coating (Scale bar ≈ 1 mm), (c) the protective polyamide sheathing around fibers (Scale bar ≈ 1 mm), (d) the finished distal tip of the probe, with the metal finger grip and single black cable containing all fibers (Scale bar ≈ 4 mm), and (e) a picture of the distal tip of the probe (Scale bar ≈ 4 mm). Please click here to view a larger version of this figure.
This multimodal instrumentation and associated technique is the first combination of these modalities within a single probe, although other combined structural/functional techniques do exist that combine different modalities. For example, hyperspectral imaging combines wide-field imaging with quantitative hemoglobin and melanin properties,17,18 and other techniques have been developed that combine optical coherence tomography (OCT) with analysis of tissue protein expression,19 to name a few. This article reports on a compact and easy-to-implement instrumentation setup that uses a general fiber-optic probe which can be optimized for various purposes including endoscopic use in the lower gastrointestinal tract and esophagus or as a handheld probe for use in the oral cavity and external skin placement.11,20
The hardware for this instrumentation requires both custom data acquisition and post-processing code to acquire diffuse reflectance spectra and then extract the resulting volume-averaged tissue physiological parameters including [Hb], [Mel], and SaO2. The custom data acquisition code was built to allow the simultaneous acquisition from a camera (for high-resolution fluorescence microscopy) and a spectrometer (for diffuse reflectance spectroscopy). Drivers are often available from the manufacturers' websites to allow integration with a variety of programming languages. The custom post-processing code imports a priori absorption values of in vivo [Hb] and [Mel]21 and then utilizes a previously developed nonlinear optimization fitting process that creates a fitted curve of the spectra.22 The fitted curve is built by minimizing the χ2 value between itself and the raw spectra and determining the tissue physiological parameters ([Hb], [Mel], and SaO2) from the fitted curve and with the lowest χ2 value.22 The code can be modified to include absorption from other chromophores as well, such as the exogenous pyranine ink used here, so that target physiological parameters are unaffected.
Physiological indicators of tissue health, such as [Hb], [Mel], and SaO2, can be used as reports of tumor response to therapy or as indicators of local vascularization and angiogenesis.14,23 Including a high-resolution fluorescence microendoscopy modality helps guide probe placement and provides investigators with a more complete picture of the relationship between epithelial tissue structure and function. In this article, construction and application of the multimodal microendoscope is described.11
Institutional Review Board approval (IRB #15-09-149) was obtained from the Human Subjects Research program at the University of Arkansas for all aspects of this study. The methods described were carried out in accordance with the approved guidelines, and informed consent was obtained from all participants.
1. Assembly of the High-resolution Fluorescence Microendoscopy Modality
Note: The outlined steps for assembly of the high-resolution fluorescence microendoscopy modality can be visualized in Figure 2.
Figure 2: Assembly of the high-resolution fluorescence microendoscopy modality. The high-resolution fluorescence microendoscopy modality can be constructed by building a shell of 1.0 inch diameter-sized components, with special care taken in handling the dichroic mirror, objective lens, excitation/emission filters, and tube lens. Glass surfaces of these components must be carefully handled using lens paper. Please click here to view a larger version of this figure.
2. Assembly of the Sub-diffuse Reflectance Spectroscopy Modality
Note: The outlined steps for assembly of the sub-diffuse reflectance spectroscopy modality can be visualized in Figure 3.
Figure 3: Assembly of the sub-diffuse reflectance spectroscopy modality. The sub-diffuse reflectance spectroscopy modality can be constructed using a basic tungsten-halogen lamp coupled to an objective lens to focus light through the 200 µm multimode delivery fiber, and a spectrometer. Additionally, a custom-built motorized optical switch can be constructed within the lamp-fiber-spectrometer path to switch between each SDS. Investigators using multiple spectrometers to acquire from multiple SDSs can bypass the optical switch component. Please click here to view a larger version of this figure.
3. Calibration of the Sub-diffuse Reflectance Spectroscopy Modality
Note: The following steps (section 3) must be completed prior to spectral data collection (section 4).
Figure 4: Calibration of the sub-diffuse reflectance spectroscopy modality. For pre-experimental calibration, the fiber-optic probe tip must be placed at different perpendicular distances from the 20% diffuse reflectance standard depending on the SDS. To consistently achieve these perpendicular distances across all experiments, a calibration standard device was designed (device cross section shown in (a)) to hold the probe at exact distances from the 20% diffuse reflectance standard. In this specific fiber-optic probe setup, light from the tungsten-halogen lamp is shown through the optical switch at source-detector separations of (b) 374 µm and (c) 730 µm (with motor and motor arm removed from the optical path for clarity). Distances of (d) 2.1 mm for the 374 µm SDS, and (e) 3.9 mm for the 730 µm SDS are required for calibration. Please click here to view a larger version of this figure.
4. In Vivo Data Acquisition and Optical Property Extraction from Human Skin
In this section, the multimodal microendoscope technique will be demonstrated on in vivo human skin.
Following this protocol, the investigator will obtain an in-focus high-resolution image of the tissue site with the full field of view (Figure 5). Outlines of cells can be seen if stained with pyranine ink from a standard yellow highlighter, whereas individual cell nuclei can be seen if stained with a dye such as proflavine. Following spectral acquisition, the post-processing software uses a priori knowledge of in vivo hemoglobin concentration ([Hb]) and melanin concentrations ([Mel])21 to fit the sub-diffuse reflectance spectra and determine values for [Hb], [Mel], and tissue oxygen saturation (SaO2) as shown in Figure 5. The post-processing software uses wide physiological bounds ([Hb] = 0-150 mg/ml, [Mel] = 0-30 mg/ml, and SaO2 = 0-100%) to fit the calibrated spectra.21
Figure 5: Co-registering qualitative and quantitative data from in vivo human normal skin and a benign melanocytic nevus. A high-resolution fluorescence image was acquired from a pyranine-ink (from a standard yellow highlighter) stained benign melanocytic nevus and adjacent normal skin tissue with an exposure time of 150 msec. Outlines of keratinocytes can are clearly visible in both images. The normal skin tissue site and melanocytic nevus had hemoglobin concentrations of 1.63 and 0.86 mg/ml, melanin concentrations of 0.78 and 10.20 mg/ml, respectively, with similar oxygen saturations of 99%. This figure demonstrates the benefit of co-registering qualitative structural and quantitative functional information. Please click here to view a larger version of this figure.
The multimodal high-resolution imaging and sub-diffuse reflectance spectroscopy fiber-bundle microendoscope reported here can be optimized and used by investigators for a variety of applications including endoscopic or handheld use for human or animal studies. It thus provides a flexible method for visualizing in vivo apical tissue micro-architecture alongside measurements of hemoglobin concentration, melanin concentration, and tissue oxygen saturation from two different tissue depths. This article describes the specifications for the fiber-optic probe, outlined a protocol for assembling the high-resolution imaging system and sub-diffuse reflectance imaging system, and shown its application in human tissues in vivo, using pyranine ink as the fluorescent contrast agent for tissue visualization. Other inks, such as proflavine or fluorescein, can be used instead of pyranine ink with appropriate approval.4-7,11
Any probe feature may be modified from this design. For the high-resolution fluorescence microendoscopy modality, the 1 mm diameter image fiber consisted of 50,000 individual core fibers with 4.5 µm spacing, resulting in a constant sub-cellular spatial resolution of 4.5 µm. Investigators wanting a different sized image fiber to obtain a smaller or larger field-of-view can find these image fibers readily available with diameters between 0.14 and 1.40 mm. A tube lens with focal length of 50 mm was chosen such that the CMOS sensor captured the full 1 mm field-of-view from the image fiber. When keeping the objective lens constant, increasing the focal length of the tube lens will increase magnification and sampling frequency but decrease field-of-view.11 Thus, the magnification of the objective lens, focal length of the tube lens, size of the image sensor, and size of the image fiber can and should be optimized depending on need. Finally, filters and excitation light source may be modified depending on the excitation/emission spectra of fluorescent dyes.4-7 In addition to modifying the probe and high-resolution fluorescence microendoscopy instrumentation, the sub-diffuse reflectance spectroscopy instrumentation can be modified.
For the sub-diffuse reflectance spectroscopy modality, different sized multimode fibers can be used at each SDS. Smaller diameter multimode fibers will be able to deliver and collect light over a smaller area, but it is recommended to use an array of identically spaced fibers to increase signal-to-noise if fiber diameters less than 200 µm are used. Investigators analyzing skin or oral tissue may benefit from an overall larger probe to increase field-of-view and signal-to-noise, but in narrower luminous organs, such as the esophagus or gastrointestinal tract, investigators will face added constraints regarding probe size, especially for compatibility with the biopsy port of conventional endoscopes.8 Other spectroscopy components that may be modified include the broadband light source and motorized optical switch. A tungsten-halogen lamp was chosen in this case, although other light sources can and have been used in other studies, including xenon arc lamps and LEDs, which may increase signal-to-noise and lower integration times.2,15,20 The motorized optical switch presented here was custom built to handle up to three SDSs, but can be modified to include more or less inputs. It should be noted that the motorized optical switch does add an additional optical component between the broadband light source and spectrometer, decreasing signal-to-noise. The switch may not be necessary for investigators with multiple spectrometers that acquire data simultaneously, but including an optical switch component ultimately reduces instrumentation cost by approximately $3,000 USD per SDS.
Construction of the instrumentation (Figures 2 and 3) is fairly straightforward. The most critical step in this protocol is the calibration of the sub-diffuse reflectance spectroscopy modality (Figure 4). Calibration must be completed immediately prior to spectral data collection. Once calibration has been completed, ensure no pieces of the instruments are shut off or re-calibration may be necessary. Proper calibration is necessary to obtain accurate reflectance spectra, and thus obtain accurate values for underlying melanin concentration, hemoglobin concentration, and tissue oxygen saturation from an unknown sample. Conveniently, most investigators use similar calibration techniques which have been well described.2,11,12,25 Information regarding software requirements for converting reflectance spectra into optical parameters can be found elsewhere.11,24,26
In regards to troubleshooting, spectra resulting in poor fits (average percent errors greater than 10% between raw data and fitted data) will yield unreliable values for the three tissue physiological parameters ([Hb], [Mel], and SaO2) presented here. Poor fits are most likely the result of either movement between the probe and skin site during data acquisition, narrow boundary conditions in the post-processing code, or unreliable a priori values of [Hb] and [Mel].11,21,24,26 Improvements in these three common error occurrences should fix the accurate fitting of sub-diffuse reflectance spectra. Thus, data collection can be improved by reducing spectrometer integration time to reduce motion artifacts within the spectra. Additionally, boundary conditions represent the range of possible computational output values for [Hb], [Mel], and SaO2 following post-processing. In these studies, boundary conditions were 0-10 mg/ml for [Hb],21,22 0-40 mg/ml for [Mel],27,28 and 0-100% for SaO2,29 which are based on values from previous studies.21,22,27-29 If measuring tissue without melanin, the lower and upper bounds for [Mel] can both simply be set to 0 mg/ml. Finally, it is recommended to use established a priori absorbance values for hemoglobin and melanin published by Prahl et al.21 These simple improvements should fix the accurate fitting of sub-diffuse reflectance spectra, and if questions remain, spectra can be validated with phantoms with known optical properties (reduced scattering and absorption coefficients).
The primary limitation to this multimodal imaging and spectroscopy fiber-bundle microendoscopy platform is the lack of a widefield imaging modality. The high-resolution fluorescence microendoscopy modality has a circular field-of-view that is 1 mm in diameter, making it difficult to rapidly scan a large area of tissue. One computational method to overcome this limitation is image mosaicking, a technique used to provide a broader field-of-view by stacking adjacent micro-scale images into a single, larger image map.10 Such image mosaicking has been previously demonstrated by Prieto et al. to investigate colonic image features.10 An instrumentation modification to overcome this limitation would be making the probe compatible with the biopsy port of a conventional endoscope, such as the probe presented by Parikh et al. to investigate colorectal neoplasia.8 This feature combines the advantages of a wide field-of-view with micro-scale imaging of high-resolution fluorescence microendoscopy.8
Overall, this technique was demonstrated on in vivo human skin and shows the value of co-registering high-resolution tissue micro-architectural images with the underlying melanin concentration, hemoglobin concentration, and tissue oxygen saturation (Figure 5). This technique can be used by researchers wishing to investigate the link between structural and functional tissue abnormalities in vivo, or analyzing tissue functional changes in the absence of observable structural changes. Future studies will investigate the viability of this technique in various epithelial disease states.
The authors have nothing to disclose.
This material is based on work supported by the National Institutes of Health (1R03-CA182052, 1R15-CA202662), the National Science Foundation Graduate Research Fellowship Program (G.G., DGE-1450079), the Arkansas Biosciences Institute, and the University of Arkansas Doctoral Academy Fellowship. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the acknowledged funding agencies.
30 mm Cage Cube with Dichroic Filter Mount | Thorlabs, Inc. | CM1-DCH | |
470 nm Dichroic Mirror (Beam Splitter) | Chroma Corporation | T470lpxr | |
Cage Assembly Rod, 1.5", 4-Pack | Thorlabs, Inc. | ER1.5-P4 | |
Cage Assembly Rod, 3.0", 4-Pack | Thorlabs, Inc. | ER3-P4 | |
Cage Assembly Rod, 2.0", 4-Pack | Thorlabs, Inc. | ER2-P4 | |
SM1-Threaded 30 mm Cage Plate | Thorlabs, Inc. | CP02 | |
SM1 Series Stress-Free Retaining Ring | Thorlabs, Inc. | SM1PRR | |
SM1 Lens Tube, 1.00" Thread Depth | Thorlabs, Inc. | SM1L10 | |
Right-Angle Kinematic Mirror Mount | Thorlabs, Inc. | KCB1 | |
1" UV Enhanced Aluminum Mirror | Thorlabs, Inc. | PF10-03-F01 | |
Z-Axis Translation Mount | Thorlabs, Inc. | SM1Z | |
10X Olympus Plan Achromatic Objective | Thorlabs, Inc. | RMS10X | |
XY Translating Lens Mount | Thorlabs, Inc. | CXY1 | |
SMA Fiber Adapter Plate with SM1 Thread | Thorlabs, Inc. | SM1SMA | |
SM1 Lens Tube, 0.50" Thread Depth | Thorlabs, Inc. | SM1L05 | |
440/40 Bandpass Filter (Excitation) | Chroma Corporation | ET440/40x | |
525/36 Bandpass Filter (Emission) | Chroma Corporation | ET525/36m | |
Quick Set Epoxy | Loctite | 1395391 | |
455 nm LED Light Housing Kit – 3-Watt | LED Supply | ALK-LH-3W-KIT | |
1" Achromatic Doublet, f=50mm | Thorlabs, Inc. | AC254-050-A | |
Flea 3 USB Monochrome Camera | Point Grey, Inc. | FL3-U3-32S2M-CS | |
0.5" Post Holder, L = 1.5" | Thorlabs, Inc. | PH1.5 | |
0.5" Optical Post, L = 4.0" | Thorlabs, Inc. | TR4 | |
Mounting Base, 1" x 2.3" x 3/8" | Thorlabs, Inc. | BA1S | |
Long Lifetime Tungsten-Halogen Light Source (Vis-NIR) | Ocean Optics | HL-2000-LL | |
20X Olympus Plan Objective | Edmund Optics, Inc. | PLN20X | |
Custom-Built Aluminum Motor Arm | N/A | N/A | Custom designed and built |
Custom-Built Aluminum Motor Arm Adaptor | N/A | N/A | Custom designed and built |
Custom-Built Aluminum Motor Housing | N/A | N/A | Custom designed and built |
Stepper Motor – 400 steps/revolution | SparkFun Electronics | ROB-10846 | Multiple suppliers |
Custom-Built Aluminum Optical Fiber Switch | N/A | N/A | Custom designed and built |
Custom-Built Aluminum Optical Fiber Switch Face-Plate | N/A | N/A | Custom designed and built |
Arduino Uno – R3 | SparkFun Electronics | DEV-11021 | Multiple suppliers |
Electronic Breadboard – Self-Adhesive | SparkFun Electronics | PRT-12002 | Multiple suppliers |
EasyDriver – Stepper Motor Driver | Sparkfun Electronics | ROB-12779 | |
12V, 229 mA Power Supply | Phihong | PSM03A | Multiple suppliers |
Enhanced Sensitivity USB Spectrometer (Vis-NIR) | Ocean Optics | USB2000+VIS-NIR-ES | |
550 µm, 0.22 NA, SMA-SMA Fiber Patch Cable | Thorlabs, Inc. | M37L01 | |
Custom-Built Fiber-Optic Probe | Myriad Fiber Imaging | N/A | |
20% Spectralon Diffuse Reflectance Standard | Labsphere, Inc. | SRS-20-010 | |
Standard Yellow Highlighter | Sharpie | 25005 | Multiple suppliers, proflavine or fluorescein can be substituted |