|Home | About | Journals | Submit | Contact Us | Français|
Optical fibers guide light between separate locations and enable new types of fluorescence imaging. Fiber-optic fluorescence imaging systems include portable handheld microscopes, flexible endoscopes well suited for imaging within hollow tissue cavities and microendoscopes that allow minimally invasive high-resolution imaging deep within tissue. A challenge in the creation of such devices is the design and integration of miniaturized optical and mechanical components. Until recently, fiber-based fluorescence imaging was mainly limited to epifluorescence and scanning confocal modalities. Two new classes of photonic crystal fiber facilitate ultrashort pulse delivery for fiber-optic two-photon fluorescence imaging. An upcoming generation of fluorescence imaging devices will be based on microfabricated device components.
Fiber-optic fluorescence imaging has become increasingly versatile over the last decade as fiber-based devices have declined in size but gained in functionality. Three classes of applications in live animal and human subjects provide the primary motivations for innovation in fiber-optic imaging. First, basic research on biological and disease processes would benefit enormously from instrumentation that permits cellular imaging under conditions in which conventional light microscopy cannot be used. Many cell types reside within hollow tissue tracts or deep within solid organs that are inaccessible to optical imaging without devices that can reach such locations in a minimally invasive manner. Flexible fiber-optic devices also allow handheld imaging and imaging in freely moving animals1. Second, fiber devices might be implanted within live subjects for long-term imaging studies. This capability will not only benefit research on the cellular effects of aging, development or experience, but also will lead to new in vivo assays for testing of drugs and therapeutics. By allowing examination of cells concurrently with observation of disease symptoms and animal behavior, fiber-optic imaging will permit studies correlating cellular properties and disease outcome in individual subjects over time and could reduce the numbers of animals needed. A third set of applications concerns development of minimally invasive clinical diagnostics and surgical procedures2. Although much work remains, fiber-optic instrumentation has already broadened the applicability of in vivo fluorescence imaging.
The seeds for development of fiber optic in vivo imaging were planted by early uses of fiber-optics and fluorescence for chemical sensing and spectroscopy, including classic work on detection of intracellular redox states3-5. Use of fiber optics for remote sensing and spectroscopy remains strong today6, and has expanded to include bioluminescence detection7. The recent proliferation of fluorescent probes has widened the set of potential uses for fiber-based fluorescence imaging devices, which may be classified by size into the three broad categories of microscopy, endoscopy and microendoscopy.
The simplest application of fiber-optics in microscopy is remote delivery of light8-16, which allows the excitation light source or photodetectors to reside apart from a conventional microscope. Optical fiber also enables miniaturized or compact forms of microscopy for portable usage1,17,18. Flexible fiber-optic endoscopy involves optical probes that are usually a few millimeters in diameter and well suited to imaging either superficial tissues such as skin or hollow tissue cavities such as the cervix or digestive tract19-34. Fluorescence microendoscopy (FME) involves optical probes that are typically 0.25–1 mm in diameter for minimally invasive insertion into solid tissue. Such probes can either be mainly rigid and one to several centimeters in length, for use akin to a needle biopsy35-37, or mated to flexible fiber optics38-43. Both types of FME have allowed cellular imaging deep within solid tissues of live mammalian subjects2,35,37,38,42,44.
All three categories of fiber-based imaging, microscopy, endoscopy and microendoscopy, can involve the fluorescence modalities reviewed elsewhere in this issue, including one-photon epifluorescence, scanning confocal fluorescence and two-photon excited fluorescence. Many strengths and limitations of the conventional tabletop systems also characterize the corresponding fiber-optic forms (Table 1). We focus here on components and embodiments that are particular to fiber-based imaging to help readers choose and construct imaging devices best suited to their needs.
Several features of fiber-optic imaging systems are nearly universal. Optical fiber delivers illumination, and one or more lenses focus this light at the specimen plane and collect fluorescence emissions. In many cases a focusing mechanism allows adjustment of the imaging plane. For confocal and two-photon modalities, a scanning mechanism moves one or more focal spots across the field of view.
Optical fiber commonly ranges from ~80 μm up to several millimeters in diameter and can perform many jobs, including light delivery and collection, as well as image transmission. How are these functions accomplished? In conventional fibers, known as step-index fibers, there are two zones with distinct indices of refraction, an outer cladding and an inner core, which enable light transmission down the fiber axis by total internal reflection45. A subclass of step-index fibers called single-mode fiber (SMF) guides only a single spatial mode of light (Box 1). SMF, which is widely available commercially, is well suited for illumination delivery in the scanning imaging modalities because the single spatial mode can be focused to a near diffraction-limited spot in the specimen plane, allowing high-resolution imaging. The same SMF core that delivers light can also act as a pinhole detector, rejecting out-of-focal plane fluorescence emissions8-11,13-15,43,46,47.
In step-index optical fibers, the fiber core has a refractive index (n1) that is typically 1–2% higher than that of the cladding (n2), leading to a numerical aperture NA = (n12 – n22)1/2 equal to the sine of the half-angle of the cone of light emitted from the fiber. The solutions to Maxwell’s equations with boundary conditions appropriate for electromagnetic wave propagation down a cylindrically symmetric fiber dictate that a fiber with a core radius a guides only a single spatial mode of light if the wavelength λ in vacuum satisfies the inequality V < 2.405, where the quantity V = (2πa NA)/λ is known as the ‘normalized wave number’45. Such fibers are single-mode and have typical core diameters of ~3–7 μm for visible light.
If V > 2.405 the fiber can guide more than one spatial mode. The number of transmitted spatial modes increases approximately quadratically as a function of V, and so-called multimode fibers operate in the limit in which the number of guided modes is large, usually a hundred or more. Multimode fibers have core diameters of ~50 μm up to several millimeters.
In cylindrically symmetric GRIN lens and fibers the refractive index commonly varies as n(r) = n0(1 – g2r2/2), where r is the radius from the fiber axis, n0 is the refractive index on axis and g is a constant parameter. In a ray description, paraxial light rays follow an approximately sinusoidal path with a period of length P = 2π/g as they propagate down the cylindrical axis of a GRIN medium45. Thus, two primary determinants of the properties of a GRIN lens are the lens length and refractive index profile.
Ultrashort pulses of ~80–250 fs duration and nanojoule energies are commonly used to induce two photon–excited fluorescence in vivo. When such pulses propagate through an optical fiber, both group velocity dispersion and self-phase modulation can distort the pulse’s temporal profile and impact the efficiency of two-photon excitation, which varies inversely with pulse duration.
Chromatic or group velocity dispersion (GVD) occurs when different colors of light travel through optical fiber or other optical components at distinct speeds, thereby incurring differential propagation delays. In optical fiber there are two sources of GVD, known as material and waveguide dispersion. Material dispersion results from the different speeds at which each spectral component travels through silica glass45.
Waveguide dispersion arises because a different proportion of the pulse energy at each wavelength propagates through the fiber cladding. These two components cancel at the so-called zero-GVD wavelength. Configurations of diffraction gratings55 or prisms56 can compensate for GVD by negatively ‘pre-chirping’ optical pulses—that is, giving wavelengths that travel more slowly through optical fiber a head start, so that all spectral components exit the fiber simultaneously. This counteracts GVD, which by itself yields pulses that are positively chirped and temporally broadened.
Intense laser pulses interact with the silica core in SMF and transiently raise the local refractive index. This index rise induces a phase shift and a corresponding spectral distortion in the pulse through self-phase modulation (SPM), which occurs in a nonlinear manner that depends on the pulse’s temporal profile of wavelength and intensity45. Because of this nonlinear dependence it is difficult to compensate for SPM, which leads to spectral narrowing and temporal broadening of negatively pre-chirped pulses propagating in SMF, for wavelengths shorter than the zero-GVD wavelength of 1.3 μm49-51. Hollow-core PBF (Fig. 1a) and LMA-PCF (Fig. 1b) ameliorate the problems associated with SPM and can deliver ultrashort pulses with little distortion.
It is possible to harness nonlinear effects in optical fiber to alter the pulse spectrum for particular two-photon imaging applications12,90,91. Small core (~1- to 3-μm diameter) or ‘highly nonlinear’ photonic crystal fibers can enhance SPM-induced spectral broadening of positively chirped ultrashort pulses below the zero-GVD wavelength (Fig. 1d). Subsequent temporal recompression of the pulses can increase the two-photon fluorescence excitation efficiency by up to almost an order of magnitude91. The spectral broadening created by SPM can also enable the excitation of dyes, such as fura-2, whose two-photon absorption spectra are below the wavelength range of most common ultrashort laser sources90.
A simple means of characterizing the temporal width of ultrashort pulses is to use an autocorrelator, which splits the beam between two paths and then recombines these on a nonlinear detector. Because the two pulses from the separate paths must arrive almost coincidentally at the nonlinear detector to register the maximum amplitude signal, pulse duration can be inferred from how the detector signal changes as the relative path length between the two paths varies. Other pulse diagnostics, such as frequency-resolved optical gating, can provide information about both the intensity and wavelength temporal profiles.
Fibers that guide more than one spatial mode are called multimode fibers (Box 1). Step-index multimode fibers are better suited for fluorescence collection than SMF, because not only do they possess larger core diameters, commonl~50 μm up to a few millimeters, but they also usually have greater numerical aperture (NA) values. All-polymer multimode fibers tend to be more flexible than their silica counterparts and are available with NA values up to ~0.5.
In a particular class of multimode fiber known as gradient refractive index (GRIN) fiber, the refractive index in the core declines approximately quadratically with radius from the central axis. Unlike in step-index multimode fibers, in GRIN fibers the different electromagnetic spatial modes comprising an image propagate at nearly the same velocity. This means that GRIN fibers can serve as lenses when cut to specific lengths (see Objective optics below), enabling remote adjustment of the focal plane in the specimen using optics placed on the opposite side of the fiber48. The number of guided spatial modes in a GRIN medium limits the optical resolution of the transmitted image. This situation contrasts with how a fiber bundle transmits a pixilated image composed of an array of light intensities, in which case an optical focal mechanism must reside on the fiber side proximal to the specimen (see Fiber bundles below).
Important issues arise in two-photon fiber-optic imaging concerning the remote delivery of the ultrashort optical pulses (~80–250 fs in duration) typically used for two-photon excitation (Box 1). Particularly for in vivo studies, it is common to use pulses of nanojoule energies for two-photon imaging up to hundreds of micrometers deep within tissue. Such brief pulses are so intense that a light-matter interaction in the glass fiber core can distort both the pulse shape and spectrum through a nonlinear process known as self-phase modulation (SPM). This effect makes it challenging to excite two-photon fluorescence efficiently using SMF49-51. A newly developed class of optical fiber known as photonic crystal or microstructured fiber can be used either to circumvent or harness SPM.
Photonic crystal fiber is named after the periodic arrays of air holes and silica glass these fibers exhibit within their internal structure (Fig. 1). Many forms of silica-core microstructured fiber guide light by a modified form of total internal reflection, in which the effective refractive index of the air-silica array in the cladding is less than that of the core. Photonic bandgap fiber (PBF) is a form of microstructured fiber that does not rely on total internal reflection for guiding light (Fig. 1a).
PBF uses diffractive effects that arise from interactions between light and internal air-silica arrays with structure at the same length scale as the wavelength. A carefully designed air-silica lattice creates a photonic bandgap analogous to an electronic bandgap in crystalline solids, which prevents light within a certain wavelength range from propagating in the cladding. An air core within such a PBF can act as a lattice defect, localizing light to the core, much as crystalline solid defects create localized electronic states. Wavelengths within the bandgap are thus contained and transmitted within the core, with the lowest-order mode typically having a Gaussian-like spatial profile52. PBF has transformed the possibilities for fiber-based two-photon excited fluorescence and other nonlinear optical imaging modalities, because even high-energy pulses do not undergo SPM in an air core. Moreover, within the transmission window of PBF, which is called the bandgap and usually covers a spectral range of several tens to over a hundred nanometers, there is usually a wavelength at which group velocity (chromatic) dispersion vanishes (Box 1). Pulses centered at this wavelength scarcely suffer any broadening, making PBF an excellent choice for pulse delivery16,17,38,53,54. For other transmitted wavelengths only tens of nanometers away, however, dispersive pulse broadening can be more severe than in conventional SMF. Application of basic methods for dispersion compensation55 can alleviate much of such broadening (Box 1), but for fiber lengths greater than a few meters additional compensation for ‘higher-order’ dispersion might also be required38,56.
A good alternative means for delivery of ultrashort pulses involves a class of microstructured fiber known as large mode area photonic crystal fiber (LMA-PCF; Fig. 1b). These low-NA fibers not only have a large core of up to ~35 μm in diameter, in which pulse intensity is much reduced compared to the ~5–6 μm core of SMF, but also are ‘endlessly single mode’ in that only one spatial mode propagates regardless of wavelength within the silica transparency window. The reduced intensity diminishes, but does not eliminate SPM. Group velocity dispersion can be compensated using basic approaches because higher order dispersion appears to be much less important than in current forms of PBF. An especially interesting LMA-PCF for two-photon imaging is a ‘double-clad’ fiber that is essentially two fibers in one and can accomplish both excitation pulse delivery and fluorescence emission collection57,58. An inner LMA core delivers ultrashort pulses, and together with an inner cladding also serves as the core for a surrounding multimode fluorescence collection fiber (Fig. 1c). An outer multimode fiber cladding is composed almost entirely of air, leading to a high NA for visible light collection up to ~0.6. Use of such double-clad fiber for two-photon microscopy has been demonstrated57,58. All types of photonic crystal fiber discussed here are commercially available.
Fiber-optic bundles consist of up to ~100,000 individual step-index fibers in a closely packed arrangement, hundreds of micrometers to a few millimeters in total diameter. Image-guiding bundles maintain the relative arrangements of the individual fibers throughout the length of the bundle, allowing transmission of an intensity image in pixilated form. Such bundles are commercially available and are commonly used for both epifluorescence59 and scanning confocal or two-photon imaging19,23,25,30,31,39,40,42,60-62. However, unlike in lenses or GRIN fibers, only intensity information is transmitted. This limitation can hinder miniaturization if one desires a built-in focusing mechanism that does not require moving the entire imaging head, because such a mechanism has to be engineered within the optics proximal to the specimen25,30,31. Many devices based on fiber bundles lack such focal capability, sacrificing functionality for size reduction39,40,42. On the other hand, use of a fiber bundle facilitates miniaturization of laserscanning modalities by allowing scanning mechanisms to reside on the side of the bundle distal to the specimen for sequential illumination of the individual fibers (Fig. 2a–c).
The main disadvantage of fiber bundles is that pixilation reduces the lateral optical resolution in the specimen plane to about twice the average core-to-core distance between fibers divided by the optical magnification of the imaging lenses. Furthermore, the thin layer of cladding between adjacent fiber cores leads to optical cross talk in which light from neighboring image pixels can leak into one another, reducing contrast. In scanning confocal imaging this problem may be reduced by use of a nonordered bundle, in which the relative fiber positions are randomized and not maintained throughout the bundle. This approach requires having to reconstruct the sample image based on knowledge of the scrambled fiber arrangement, but helps to reduce cross talk. Nonordered bundles are especially useful for laser line-scanning forms of fast confocal imaging, in which adjacent fibers at one side of the bundle are illuminated concurrently but the resulting fluorescence originates from physically separate locations in the sample62. In all fiber-bundle applications, calibration of the intrinsic autofluorescence, precise position and optical throughput for each fiber in the bundle can facilitate postacquisition image processing and substantially improve image quality42.
For applications without severe size constraints, a conventional microscope objective lens nearly always provides superior optical resolution and fluorescence detection efficiency. When miniaturization is required, custom high-NA (0.46–1.0)19,23,30,31 objective lenses can be designed with outer diameters as small as 3–7 mm. Use of injection-molded plastic lenses to construct a custom objective is a particularly lightweight and economical approach23.
The creation of minimally invasive microendoscope probes has benefited greatly from recent advances in the fabrication of GRIN microoptics. GRIN microlenses ranging from 350–1,000 μm in diameter have been used for one-photon, confocal and two-photon FME35-40,43. As in GRIN fiber, cylindrical GRIN lenses exhibit a refractive index that declines approximately quadratically with radius (Box 1). Compound GRIN microendoscope probes usually measure one or more centimeters in length and combine two or more GRIN lenses with different characteristics35-37. These are typically a 0.4–0.6 NA objective lens that enables micrometer-scale resolution and a longer relay lens that projects a real image of the specimen plane and supplies sufficient length for reaching deep tissue35-37. Optical aberrations in current GRIN lenses limit the resolving power to approximately twice the diffraction limit35-38. A detailed description of GRIN endoscope probes is available in ref. 35, and custom GRIN lenses are available from multiple vendors.
It is common to speak of scanning mechanisms as being either proximal or distal to the light source. Proximal scanners are located in the illumination pathway upstream of the fiber and are used with a fiber bundle. Distal scanners are located on the fiber side distal to the light source and usually scan illumination from a single fiber over the specimen. Proximal scanning offers the benefit of separating bulky scanners from a miniaturized imaging head and typically involves a pair of galvanometer-mounted scanning mirrors (Fig. 2a)19,23,25,30,31,39,40,42,61,62. These can provide high image-acquisition rates (up to ~1 kHz per image line), particularly in a fast line-scanning approach that can allow video-rate imaging (Fig. 2b), and optical deflection angles up to ~40°. However, because the galvanometer approach involves sweeping the illumination across the bundle, light enters the cladding between adjacent cores during much of the scanning cycle, potentially diminishing image contrast.
Use of a spatial light modulator such as a pixilated micromirror array to illuminate each fiber sequentially, without sweeping the beam, can reduce errant excitation light (Fig. 2c)60. Nonetheless, much of the excitation power is lost, because although the entire modulator is illuminated with light, only a fraction of the pixels will be activated at any given moment during a scan. This loss may not pose much problem for imaging at superficial tissue depths. Both one-dimensional (1D) and two-dimensional (2D) spatial light modulators are available commercially.
The use of a single fiber for illumination delivery necessitates a distal scanning mechanism. Portable microscopy can rely on a piezoelectric-driven, 2D tip-tilt mirror (Fig. 2d), which has a lower bandwidth (~200 Hz) but a smaller footprint than cascaded galvanometers17,18,63. However, endoscopy and microendoscopy require more compact scanning mechanisms. One approach relies on either a piezoelectric1,38,64 or an electromagnetically excited65,66 actuator to drive mechanical resonance vibrations of the fiber tip (Fig. 2e). Although fiber tip scanning allows stationary imaging optics to be inserted into solid tissue38, it can also introduce off-axis aberrations that might limit the achievable resolution. To preclude such aberrations, one can attach both the fiber and the imaging lens to a resonant cantilever and scan them together using electrostatic forces67 or piezoelectric actuators20,24 (Fig. 2f). Although this method needs an actuator that delivers more force, the required angular scanning range is reduced because the scanned focal spot is not further demagnified. A disadvantage of resonant scanning is that it is usually not possible to offset the center of the image field from the main optical axis without addition of another deflector.
In recent years, microfabricated microelectromechanical systems (MEMS) torsion scanning mirrors ~0.5–2 mm in diameter have emerged as versatile miniaturized scanning mechanisms (Figs. 2d2d and 33)68-70. MEMS scanners are created through sequential material etching and deposition processes analogous to those used for integrated circuit fabrication; 2D scanning involves cascaded 1D scanners or a single 2D scanner. The latter simplifies the optical design and reduces overall size. MEMS scanners that rely on electrostatic actuation are the most commonly used owing to the small size, low power consumption and high force that can be achieved. Among the electrostatic actuators, vertical comb actuators71-73 offer greater force and angular range than parallel plate actuators68,69. These interdigitated silicon comb actuators provide electrostatic torque to rotate the mirror in one direction, and torsional springs supply a restoring torque (Fig. 3). Operating such scanners in DC mode permits static offsets of the field of view and mechanical angular rotations of up to ~6° (ref. 71). Alternatively, because ~1-mm-diameter MEMS mirrors have mechanical resonance frequencies in the kHz range, resonance scanning can enable videorate imaging with large angular mirror rotations (>30°)72, albeit at the expense of not being able to make lateral offsets in the field of view. A disadvantage of MEMS scanners is that although they are ideal for miniaturization, their fabrication requires much expertise. However, MEMS scanning mirrors are becoming commercially available.
Many applications benefit from having a built-in mechanism for altering the image focus without having to move the entire imaging probe. As with scanners, such mechanisms are positioned either proximal or distal to the light source relative to the fiber. Distal focusing mechanisms tend to hinder miniaturization because they are integrated with the objective optics, whereas proximal mechanisms are of conventional size and do not enter the specimen. For instance, it can be helpful to leave a GRIN microendoscope probe stationary within tissue while adjusting the image plane by moving the external microscope objective lens that also delivers illumination into the probe35,37,74. Because of optical demagnification between image and object planes, a focal shift of ~100 micrometers in the tissue often requires movement of focusing optics over millimeter-sized distances, beyond the range of small piezoelectric actuators. For flexible modalities that require distal focusing, one can use a hydraulic or pneumatic system to move the end of a fiber with respect to the objective optics when a fluid is injected into or withdrawn from a reservoir25,30,31. Another lightweight approach relies on a miniature motor to alter the distance between the fiber and the imaging optics38. In the future, millimeter-sized liquid lenses that alter their focus via electrowetting, will allow fast, remotely controlled distal focusing and beam steering for endoscopy and FME75-77.
As imaging systems have become increasingly miniaturized, optical and mechanical designs have grown increasingly intricate. Zemax and Code V are optical ray-tracing software programs that are commonly used for optical design and tolerance analysis. Computer-aided design programs such as SolidWorks and AutoCad can generate 3D models and technical drawings of mechanical configurations. We have found such software invaluable for designing complex imaging systems.
Fiber-optic epifluorescence, scanning confocal and two-photon imaging each has distinct sets of strengths and limitations78 (Table 1). The latter two forms offer 3D optical sectioning, important for many applications, but sectioning is not always needed or desired. For example, in vivo epifluorescence microendoscopy readily allows viewing, by eye or with video-rate recordings, of red blood cell flow, without the need for laser sources or fast scanners35,44. Further, lack of sectioning may facilitate faster imaging speeds in some applications by aggregating fluorescence signals acquired simultaneously from adjacent tissue depths. The key advantage of two-photon imaging is robustness to scattering and increased penetration depths, but the high cost of an ultrashort-pulse laser is a distinct disadvantage. Nonetheless, when using epifluorescence or confocal modalities to image through more than ~50 μm of tissue, image contrast declines sharply with depth due to scattering. Additionally, these two modalities are more vulnerable to photobleaching and phototoxicity, which unlike in two-photon imaging occur not only at the focus but also in out-of-focus planes.
One-photon epifluorescence imaging can involve a compound GRIN microendoscope probe, a flexible fiber bundle or both. Unlike existing forms of either fiber-optic confocal or two-photon imaging, one-photon endoscopy and microendoscopy are currently the only fiber imaging varieties that enable images to be directly viewed by eye.
In the simplest form of one-photon FME (Fig. 4a), a conventional microscope is equipped with a GRIN endoscope probe ≤1 mm in diameter. The microscope objective directs illumination light into the probe and adjustment of the objective position provides a noninvasive focusing mechanism for altering the image plane in the tissue35,36. This approach has enabled imaging of neuronal activity and blood flow in deep brain regions of live rodents2,35,79, but the use of short, rigid microendoscope probes coupled to a full-sized microscope restricts applications to immobilized specimens.
The use of fiber bundles alleviates this stipulation and provides mechanical flexibility while imaging in deep tissue59 (Fig. 4b). Similar approaches have been used for reflectance imaging in freely moving cats80. With a bare fiber bundle, only objects in direct contact with the tip of the bundle are in focus, but addition of a miniature lens to the distal end of the bundle can add optical working distance. As with conventional epifluorescence microscopy, out-of-focus and scattered fluorescence photons reduce image contrast.
Fiber-optic confocal imaging can involve one SMF, two SMFs for dual-axis imaging or a fiber bundle. SMFs are used in confocal microscopy for remote light delivery, but their main application is in flexible endoscopy. Miniaturized optics and compact distal scanning mechanisms, such as fiber-tip scanning22,65 (Fig. 2e), fiber-objective lens scanning20,24,67 (Fig. 2f) or MEMS mirror–based scanning (Figs. 2d2d and and33)68,69, have enabled SMF-based confocal endoscopes with an outer diameter as small as 3.5 mm (Fig. 4c)24. Nonetheless, reduction of imaging and scanning optics to the millimeter scale nearly always compromises some combination of resolution, field of view and working distance. With confocal endoscopy based on one SMF (Fig. 4c), design pressures tend to center on the choice of high NA objective optics, which are a key determinant of image quality and fluorescence collection efficiency.
Dual-axis confocal imaging provides an alternate set of compromises, reducing ease of alignment and efficiency of fluorescence collection to gain longer optical working distance, immunity to off-axis aberrations, and good axial resolution using economical objective lenses. Two SMFs mounted at an angle to one another deliver light and collect fluorescence, respectively (Fig. 4d). Axial resolution is determined by the angle between the fibers and the transverse size of the illumination beam. Thus, two low-NA, long-working-distance objectives can provide an effective high-NA for axial sectioning, although the fluorescence collection NA remains low. A compact postobjective scanning mechanism, such as a 2D tip-tilt mirror (Fig. 2d), offers broad fields of view up to ~500 μm81-83. To date, dual-axis imaging has been demonstrated in microscopy, but dual-axis endoscopy should emerge soon.
Fiber bundle confocal imaging facilitates miniaturization by use of conventional scanners proximal to the light source but reduces resolution through image pixilation (Fig. 4e). Many fiber bundle confocal designs also forgo a focusing mechanism to reduce size. For example, simply fusing a compound GRIN micro-lens onto the tip of a bundle enables handheld confocal FME probes with diameters as small as 300 μm40,42. Such probes and accompanying imaging systems are now commercially available. By comparison, fiber bundle confocal endoscopes with remotely controlled distal focusing and miniature objective lenses exist with outer diameters as small as 3 mm30. With either variety, a descanning pinhole rejects out-of-focus fluorescence to achieve optical sectioning. Scanning a line of excitation light across the bundle (Fig. 2b) boosts the image-acquisition rate, enabling video-rate imaging, although one must then use a descanning slit that does not reject out-of-focus light along the slit’s linear dimension25,30,31.
For two-photon FME, a conventional two-photon microscope can be furnished with a GRIN microendoscope probe for insertion into tissue (Fig. 4f). As in one-photon FME, adjustment of the microscope objective position alters image focus. Although image frame rates have been lower (up to several Hz) than in one-photon FME, localized two-photon excitation provides inherent optical sectioning up to hundreds of micrometers from the tip of the endoscope probe. Two-photon FME has enabled in vivo imaging of individual neurons and blood vessels deep in the brain35,37.
As with one-photon endoscopy, the combination of a short endoscope probe and a flexible fiber removes the requirement for specimens to be immobilized (Fig. 4g). A two-photon FME system based on a PBF for delivery of ultrashort pulses and a GRIN microendoscope probe has been created with an imaging head that is ~3.5 cm × 1.2 cm × 1.5 cm in size and 3.9 g in mass38. This device uses resonant fiber tip scanning1, DC micromotor–based focusing and collection of fluorescence using a large-core multimode fiber. By comparison, fiber-optic two-photon microscopy based on a small microscope objective provides finer resolution and higher collection efficiency but does not allow the same degree of miniaturization1.
For both flexibility and superior miniaturization, a fiber bundle based approach to two-photon FME has been realized (Fig. 4h)39. Because fluorescence excitation is localized to the focal spot, the descanning pinhole of confocal modalities is unnecessary, enabling whole-area detection in which fluorescence is collected across the entire bundle. Nonlinear distortion of the excitation pulses in each fiber (Box 1) currently limits fiber-bundle two-photon imaging to applications in surface tissue that do not require nanojoule energy pulses39. Unfortunately, arrays of hollow core PBF or LMA-PCF do not yet exist with sufficiently large numbers of pixels for imaging.
The combination of a single PBF for light delivery, a micro-lens array for generating multiple foci and a small 2D scanning mirror has enabled high-speed, handheld two-photon microscopy (Fig. 4i)17,18. Because multiple laser foci exist within the sample it is necessary to use a camera or multianode photomultiplier tube for signal detection. When the field of view is small and imaging is performed deep within tissue, scattering of fluorescence emissions can lead to cross talk between pixels and reduced contrast. To date, multifocal, two-photon imaging has been demonstrated for portable microscopy17,18, but further miniaturization might enable endoscopy.
Fiber-optic fluorescence imaging will continue to benefit from progress in miniaturized and microfabricated components. There is also growing interest in other nonlinear optical modalities, such as second harmonic generation (SHG)84 and Coherent Anti-Stokes Raman Scattering (CARS)85, that provide inherent optical sectioning as in two-photon imaging but with intrinsic contrast generation mechanisms that do not require a dye label. Fiber-optic SHG and CARS imaging will benefit from the same trends towards miniaturization that are driving progress in fiber-optic fluorescence imaging.
Future fiber-optic systems will likely use microfabricated components in addition to scanning mirrors. Electrostatically actuated lenses are an alternative to MEMS mirrors for miniaturized scanning86. In addition, optical fibers and small lenses can be integrated by micromachining a v-groove or optical benches directly into a silicon wafer68,87,88. Optical fibers for light delivery and collection may eventually become unnecessary with the advent of integrated semiconductor laser sources and photodetectors89. Imaging systems in which all components are fabricated monolithically would provide a new degree of miniaturization, inherent alignment and potential for highly parallel imaging. Together, such advances will propel imaging applications that require both portability and a high degree of functionality.
Our work on fiber-optic imaging is supported by grants to M.J.S from the Human Frontier Science Program, the US National Institute on Drug Abuse, the US National Institute of Neurological Disorders and Stroke, the US National Science Foundation, the US Office of Naval Research, the Arnold & Mabel Beckman Foundation and the David & Lucille Packard Foundation. B.A.F is a National Science Foundation Graduate Research Fellow. E.D.C. is a member of the Stanford Biotechnology Program. W.P. is an affiliate of the National Electronics and Computer Technology Center of Thailand. E.L.M.C. is supported in part by a Dean’s Fellowship, Stanford School of Medicine. We thank R. Kristiansen of C. Fibre for providing images of photonic crystal fiber.
COMPETING INTERESTS STATEMENT
The authors declare that they have no competing financial interests.