2.1.

Longitudinal and Transverse Resolution

In contrast to conventional and confocal microscopy, OCT achieves very high axial image resolution independent of focusing conditions because the axial and transverse resolutions are determined independently. The transverse resolution as well as depth of focus are governed by the focal spot size (as in conventional microscopy), while the axial resolution is mainly governed by the coherence length of the light source (rather than the depth of field, as in microscopy). The coherence length is the spatial width of the field autocorrelation measured by the interferometer. The envelope of the field correlation is equivalent to the Fourier transform of the power spectrum. The axial resolution is therefore inversely proportional to the spectral bandwidth of the light source, to Δz=2 ln(2)λ²/(πΔλ), where Δλ is the bandwidth and λ is the center wavelength of the light source used for imaging (assuming a Gaussian spectrum).

Figure 1 depicts this correspondence for the two standard wavelength regions used for OCT so far (800 nm and 1300 nm), assuming a Gaussian spectrum of the light source as well as a nondispersive imaging medium. It indicates that by using a light source with a 140-nm optical bandwidth at an 800-nm center wavelength, a free space resolution of approximately 2 μm can theoretically be achieved. It also indicates that for a further improvement by another factor of 2, down to 1 μm, the bandwidth of the light source must be doubled to 280 nm. This makes it very hard to achieve a longitudinal resolution in OCT imaging of less than 2 μm, because very broad bandwidth light sources are required. Figure 1 also shows that using a bandwidth of 180 nm centered at 800 nm enables a free space resolution of 1.6 μm. However, the same amount of bandwidth centered at 1300 nm enables only 4.1 μm. Hence even broader bandwidth light sources have to be used when employing light sources in the 1300-nm wavelength region. The actual resolution within the imaged tissue can be estimated by dividing the free space resolution, indicated by this figure, by the group refractive index, i.e., 1.35 to 1.4 for most biological tissues. It has to be noted that the axial resolution of OCT imaging is limited by dispersion of the sample, but also absorption and scattering within the sample, where photons with the same path but reflected from various imaging depths are detected.

Figure 1

Theoretical calculation of free space resolution (Δz at full width half maximum) as a function of the two standard wavelengths (λ=800 and 1300 nm) used for OCT and optical bandwidth (Δλ). A Gaussian spectrum and nondispersive medium are assumed.

An in conventional microscopy, the transverse resolution for OCT imaging is determined by the focused transverse spot size (defined as the 1/e² radius of a Gaussian beam) of the optical beam, given by Δx=(4λ/π)(f/d) where d is the spot size on the objective lens and f is its focal length. The transverse resolution is also related to the depth of focus or the confocal parameter 2z_R (two times the Rayleigh range): 2z_R=πΔx²/2λ. This implies that increasing the numerical aperture reduces the spot size, but also reduces depth of focus. High transverse-resolution OCT imaging can be achieved by focusing with a high numerical aperture at the expense of reduced depth of focus. This operating regime is similar to conventional microscopy or confocal microscopy, which has high transverse resolution but poor depth of field; for example, with a transverse resolution of 3 to 5 μm using a light source centered at 800 nm, a corresponding confocal parameter of only 35 to 100 μm in air can be achieved. To overcome this depth of field limitation caused by the high numerical aperture, C-mode scan imaging,⁹⁵ the use of special imaging lenses,⁹⁶ or dynamic focus tracking^{97 98} can be used to maintain high transverse resolution throughout the whole penetration depth.

Especially for ophthalmic retinal OCT imaging, low numerical aperture focusing is employed because it is desirable to have a large depth of field and to use OCT to achieve high axial resolution. For a center wavelength of λ=800 nm, a beam diameter of 2 to 3 mm, and neglecting the chromatic aberration of the human eye, the depth of field of the focus is approximately 750 μm, which is comparable to the depth over which retinal imaging is performed. However, low numerical aperture focusing yields a low transverse resolution. Studies have shown that the largest pupil size that still yields diffraction-limited focusing is approximately 3 mm, enabling theoretical retinal spot sizes of 10 to 15 μm.⁹⁹ The minimum achievable transverse resolution for OCT is determined by the smallest achievable spot size on the retina. Using a reduced schematic eye for the optical analysis and diffraction-limited focusing of the eye, a 4-mm diameter beam at the pupil (with a center wavelength of 800 nm) would produce a theoretical spot size of 4.3 μm. In practice, however, for large pupil diameters, ocular aberrations limit this minimum focused spot size on the retina, even for monochromatic illumination. When ultrabroad bandwidth light is used, chromatic aberration will pose an additional limit to the smallest possible spot size. Furthermore, coma and other third-order aberrations, rather than spherical or fourth-order aberrations, appear to place the dominant limitation on the resolving power of the eye. One promising approach to correcting ocular aberrations in order to decrease the spot size at the human eye fundus and to therefore improve transverse resolution in OCT would be to use adaptive optics (AO).

The applications of AO have gone far beyond astronomy in the past decades. Correcting ocular aberrations was suggested in the early 1960s. With advances in technology, the first promising results were achieved in the late 1980s.¹⁰⁰ Today, deformable mirrors, liquid-crystal spatial light modulators, and phase plates have the potential not only for static, but also for dynamic correction of ocular aberrations.^{101 102} Integrating this technique into ultrahigh axial-resolution ophthalmic OCT would enable even further improvement of the visualization possibilities of ophthalmic OCT. Combining AO with ultrahigh-resolution ophthalmic OCT should—besides providing an axial resolution of about 3 μm—enable a transverse resolution of about 5 μm. In addition, proper focusing onto the retina should also improve the sensitivity of the system through a more efficient collection of backscattered light.

2.2.

Dynamic Range, Sensitivity—Data Acquisition Speed

OCT can achieve high detection sensitivity because interferometry measures the field rather than the intensity of light. Interferometric detection is a powerful approach for measuring the echo time delay of backscattered light with high sensitivity. This technique is closely related to heterodyne detection in optical communications. Weak backscattered optical signals are effectively multiplied by the reference signal. The signal-to-noise (SNR) performance of detection can be calculated from optical communication theory. The signal-to-noise performance is given as SNR=10 lg(ηP)/(E_pNEB), where P is the power that is reflected or backscattered, NEB, the noise equivalent power, is the bandwidth of the electronic detection, η is the photodetector efficiency, and E_p is the energy of the photon. This formula essentially states that the signal-to-noise performance is proportional to the power detected divided by the bandwidth measurement. This bandwidth is not the optical bandwidth of the light source, but instead refers to the electronic bandwidth or data acquisition rate. This formula also shows that there is a tradeoff between sensitivity, speed, and resolution. Furthermore, a minimum optical output of the light source employed is necessary in combination with an optimum system design^{103 104} to enable near quantum-limited detection.

The specifications of optical coherence tomography systems vary widely according to their design and data acquisition speed requirements. A major limiting factor—especially for ultrahigh-resolution ophthalmic OCT—in dynamic range and sensitivity is the fact that the maximum allowed illumination power is limited by tissue or retinal exposure considerations. Superluminescent diodes, in contrast to femtosecond solid-state lasers, have the advantage of emitting laser light with extremely low-amplitude noise. In contrast, solid-state femtosecond lasers have excess amplitude noise that can decrease the signal-to-noise ratio and therefore the sensitivity of the system. Nevertheless, optimum imaging performance can still be obtained with these light sources to achieve nearly shot-noise-limited sensitivity by using dual-balanced detection.¹⁰⁵ Hence, typical OCT systems achieve detection sensitivities of 10⁻⁹ or 10⁻¹⁰ (−90 to −100 dB) of the incident optical power. This is sufficient to detect optical signals from depths of 1 to 2 mm in most scattering biological tissues.

Different approaches have been used so far to accomplish high-speed, real-time OCT imaging. The use of a rapid-scanning optical delay line¹⁰⁶ enabled in vivo video rate OCT imaging.¹⁰⁷ En face OCT uses rapid transverse scanning to achieve high-speed OCT imaging with large voxels per second acquisition rates.¹⁰⁸ Two other alternatives for high-speed OCT imaging are parallel OCT using a smart detector array¹⁰⁹ or Fourier domain OCT.^{110 111}

2.3.

Center Wavelength, Penetration, and Image Contrast

OCT imaging at different wavelengths can be used to enhance tissue contrast and penetration, as well as to measure absorbing or scattering properties of various pigments and structures. Tissue scattering and absorption properties are strongly dependent upon tissue morphology, such as cell and nuclei size, shape, and density, as well as the physiological state of certain chromophores. Hence, OCT penetration depth is significantly affected by light scattering within biological tissue, which is scaled as 1/λ_c ^k, where the coefficient k is dependent on the size, shape, and relative refractive index of the scattering particles.⁸⁹ The difference in tissue scattering and absorption provides structural contrast for OCT. Image contrast can be enhanced by the use of polarization-sensitive OCT,^{86 87 88} phase-contrast OCT,^{112 113} or by using spectroscopic OCT. Most biological tissues absorb in the visible and near-IR wavelength range because of the presence of hemoglobin and melanin. In the range between 0.8 and 1.8 μm, scattering is the predominant mechanism limiting image penetration depth. Since scattering depends strongly on wavelength and decreases for longer wavelengths, significantly better image penetration depth can be achieved with light centered at 1.35 μm than at 0.85 μm. It has been shown that the optimum wavelengths for imaging in biological tissues are in the 1.3 to 1.5-μm range. In this range imaging depths of 1 to 2 mm can be achieved.^{114 115} Water absorption becomes a problem for wavelengths >1.8 μm. A strong water absorption peak at ∼1.43 μm can also modulate the spectral width of a broad bandwidth light source centered at 1.3 μm, as a function of depth in biological tissue, since most soft tissues have 50 to 90 water content. Hence, if the center wavelength and bandwidth of the coherent light source used for OCT could be chosen across the range from 500 to 1500 nm, axial resolution and contrast as well as penetration could be optimized according to the imaging procedure and medical application.

3.1.

Spectral Shape of Ultrabroad Bandwidth Light Sources

The necessary light source specifications for ultrahigh-resolution OCT are not only spatially coherent, ultrabroad bandwidth emission with enough output power and good amplitude stability, but also optimum spectral shape. Since the coherence length is defined as the full width at half maximum (FWHM) of the field autocorrelation measured by the OCT interferometer, the width and also the shape of the coherence function of an OCT system is dependent on the spectral shape of the light source as well as on the transfer function of the OCT system. The latter is mainly determined by the optical properties of the interferometer, e.g., wavelength-dependent losses and splitting ratios of beamsplitter and lenses, as well as the cutoff wavelength of the single-mode fibers employed and the wavelength-dependent sensitivity of the photodetectors employed. The ideal spectrum for OCT would have a Gaussian spectral shape, resulting in a Gaussian coherence function with no side lobes. Large spectral modulations would reduce sensitivity and resolution, owing to the presence of side lobes in the fringe pattern that appear symmetric to the maximum of the coherence functions.

There are several possibilities for changing the shape of the emission spectrum of the light source used. The easiest way is to introduce an optical dichroic or interference filter that suppress certain wavelength regions. Another possibility is to spatially disperse the optical beam with prisms and to induce local and therefore wavelength-dependent losses by introducing razor blades or thin objects into the dispersed light beam. Figure 2(a) demonstrates the effect of this shaping method when applied to a strongly modulated spectrum of an ultrabroad bandwidth titanium:sapphire (Ti:sapphire) laser. Alternatively, the configuration of this prism can be used in combination with a liquid-crystal spectral amplitude or phase shaper (spatial light modulator), a well-established technique for pulse-shaping applications in ultrafast optics.¹¹⁶ With this technique, selected wavelength regions of less than 10 nm can be attenuated by a factor of up to 80. Figures 2(b) and 2(c) demonstrate the capabilities of this technique. Note that this technique is rather expensive and more challenging concerning the control of the 640 liquid-crystal elements. In addition, spectral shapers introduce power losses of up to 50. In the case of light sources that are based on nonlinear optical effects in special fibers, the proper choice of the input pulse parameters into the fiber can be used for spectral shaping of the output spectrum generated.¹¹⁷ Figure 2(d) depicts an example where the fiber length that has been used to generate a supercontinuum has been optimized to minimize spectral modulations. By carefully optimizing the pulse input parameters, emission spectra with up to 325 nm could be generated with less than 1.5 dB modulations directly out of the fiber, with no additional spectral shaping (cf. Sect. 4.3).

Figure 2

Spectral shaping methods; black lines indicate original spectrum, gray lines indicate shaped spectrum. (a) Shaping by introducing wavelength-dependent attenuation. (b) and (c) Effect of liquid-crystal-based spatial light modulator. (d) Optimizing input pulse parameters in supercontinuum generation with photonic crystal fibers (PCF); in this case the PCF length has been optimized.

3.2.

Chromatic Aberration Limitations to Resolution

Another limitation for achieving ultrahigh resolution is the chromatic aberration of the optics employed. Conventional lenses have a focal length that varies with wavelength and thus they focus ultrabroad bandwidth light an different focal positions. This variation in focal position for different wavelengths limits the actually effective bandwidth and therefore degrades resolution. Figure 3 depicts a comparison of spectra at the exit of an ultrahigh-resolution OCT interferometer for different focal distances when standard—uncorrected—optics [Fig. 3(a)] and specially designed optics with minimal chromatic aberrations [Fig. 3(b)] is used in the sample arm. It is obvious that for different focal distances of the uncorrected optics [Fig. 3(a)], a different actual spectrum is effective, resulting in different actually effective bandwidths at full width at half maximum and therefore a spatially changing axial resolution. In the case of a specially corrected achromatic optics, different focal distances just introduce a wavelength-independent attenuation of the whole spectrum, maintaining the spectral shape, optical bandwidth, and therefore axial resolution. Hence, appropriately coated and designed achromatic objectives have to be used to maintain the ultrabroad bandwidth of the light in order to achieve ultrahigh resolution. Alternatives are catadioptic (reflective) objectives of parabolic mirrors that have minimal chromatic aberration and do not introduce any dispersion. For ophthalmic OCT, since the human eye is not a diffraction-limited optical system, chromatic aberration in the eye itself may ultimately limit the axial resolution of ultrahigh resolution retinal OCT imaging.

Figure 3

Effect of chromatic aberration. Optical spectra at the exit of an OCT interferometer. Spectra are displayed as a function of focusing distance onto a mirror in the sample arm, employing (a) a standard (not achromatic) as well as (b) an optimized achromatic lens in the sample arm. Standard lenses introduce spatially changing effective optical bandwidths and therefore spatially changing axial resolution.

3.3.

Group Velocity Dispersion Limitations to Resolution

As mentioned before, the axial resolution of OCT is mainly determined by the bandwidth of the low-coherence light source. However, chromatic (or group velocity) dispersion is another limiting factor that degrades resolution by broadening the axial point spread function. Chromatic or group velocity dispersion causes different frequencies in the broad bandwidth light to propagate with different velocities. It should be noted that the group velocity dispersion changes the phase but not the bandwidth of optical signals. The axial point spread function, i.e., the interferometric autocorrelation function, will also broaden and therefore decrease resolution if there is a significant mismatch in the dispersion of group velocity between the reference and the sample arms of the measurement system. Both the fiber optics and the sample itself (especially in the case of the human eye) may have significant dispersion. The effect of broadening will become more severe with increasing optical bandwidth and thickness of the sample and is less pronounced for longer wavelengths. A dispersion mismatch between the interferometer arms of the OCT system also degrades the peak height of the interference envelope, which reduces the system’s dynamic range.

In order to compensate for these effects, dispersion can be compensated by using a numerical a posteriori depth-variant dispersion compensation, or dispersion introduced by the fiber length and optics mismatch between the sample and the reference arm can be carefully balanced by using variable thicknesses of fused silica (FS) and BK7. This dispersion balancing can be done in real time by performing fast Fourier transform (FFT) analysis of the interference signal of a single reflecting site sample. The thickness of fused silica and BK7 in the reference arm can be changed systematically until a uniform group delay dispersion, i.e., the first derivative of the phase that has been obtained by FFT, is achieved. Figure 4 illustrates this real-time dispersion balancing technique. It shows interference fringes detected when using a single reflecting sample [e.g., a mirror Fig. 4(a)], corresponding spectrum [Fig. 4(b)], the derived group delay [Fig. 4(c)], and group delay dispersion [Fig. 4(d)]. For ultrahigh-resolution retinal OCT imaging, the dispersion of approximately 25 mm of ocular media can be compensated by using 25 mm of water in the reference arm, since previous in vivo studies of dispersion measurements have shown that the dispersion of ocular media averaged over 25 mm is similar to that of 25 mm of water.⁴¹

Figure 4

Online OCT system dispersion compensation. (a) Interference fringe signal detected from a mirror in the sample arm. (b) Spectrum obtained by Fourier transformation. (c) Group delay from the first derivative from the phase, indicating a dispersion match or mismatch. (d) Group delay dispersion from the second derivative of the phase, indicating a higher-order dispersion mismatch.

Balancing the higher-order dispersion distribution of the 25-mm ocular media in front of the retina is more critical for obtaining high resolution than balancing the dispersion of the system itself. Figure 5 and Fig. 6 demonstrate this effect. The dispersion-balanced [Fig. 5(a)] and unbalanced [Fig. 5(b)] interference signal of double-passing 25 mm of water is depicted in Fig. 5. Resolution degrades from 4 μm to 52 μm, or more than a factor of 10. Figure 6 depicts the effect of dispersion mismatch in the case of in vivo ultrahigh-resolution retinal imaging of a normal human fovea [Fig. 6(a)]. By artificially introducing dispersion mismatch by inserting 3 mm [Fig. 6(b)] and 9 mm [Fig. 6(c)] of fused silica in the reference arm of the OCT interferometer, it is obvious that both OCT resolution and sensitivity (up to −5 dB) decrease. Clinical application of ultrahigh-resolution ophthalmic OCT revealed that by using a dispersion compensation of 25 mm of water in the reference arm of the interferometer, patients with an axial eye length between 23 mm and about 27 mm can be imaged without any significant loss of axial resolution induced by a dispersion mismatch.

Figure 5

Dispersion balanced (a) and unbalanced (b) interference signal after double-passing 25 mm of water, introducing a dispersion that is the same thickness as ocular media. Resolution degrades by more than 10 times.

Figure 6

Effect of dispersion mismatch in in vivo ultrahigh-resolution retinal imaging. (a) Dispersion matched. (b) Artificially introduced dispersion mismatch by 3 mm (middle) as well as (b) 9 mm of fused silica in the reference arm. Clear axial resolution as well as a degradation in sensitivity (up to −5 dB) is observed.

3.4.

Other Limitations to Resolution

Another effect that limits axial OCT resolution in ultrahigh-resolution OCT systems is a polarization mismatch between the interferometer arms, which introduces a phase difference and therefore a change in the shape of the coherence function and axial resolution, respectively.

The optical transmittance, coating, and wavelength-dependent losses of the bulk or fiber optics employed, the delivery system optics, and the human eye itself—in the case of ultrahigh-resolution ophthalmic OCT—strongly influence the axial resolution as well as the sensitivity of the OCT system. Single-mode fibers with appropriate cutoff wavelengths have to be used to provide single-mode light propagation. Conventional fiber couplers are designed to maintain 3-dB splitting over a wavelength range of typically ±10 nm. By using these fiber couplers for delivery of broad-bandwidth laser light, unequal beam splitting with respect to wavelength and power can occur and will reduce resolution. Hence special broad-bandwidth, wavelength-flattened, 3-dB fiber or bulk optic beamsplitters have to be used to maintain broad bandwidths and consequently high axial resolution. It is important to note that bidirectionality of bulk or fiber-based beamsplitters is important in order not to introduce wavelength-dependent losses on the way back to the detector.

The detection system, including the electronics as well as the digitization and acquisition of the interference signal, have to be designed properly to avoid degradation of axial OCT resolution. Hence, the transimpedance amplifier, but more particularly the electronic bandpass filtering, must be designed properly and adapted to the ultrabroad optical bandwidth. The electronic bandwidth of the bandpass must not be so narrow as to reduce the axial resolution, but must also not be so broad that it reduces sensitivity by introducing noise. Real-time adaptive filtering might be a proper approach to optimize sensitivity and maintain axial resolution. In addition, hardware demodulation must be adapted to the scanning speed and optical bandwidth to avoid broadening caused by the time response of the electronics used, resulting in a larger coherence length of the envelope compared with to the full interference fringe signal. Finally, the signal must be temporally digitized correctly with at least 5 to 10 times oversampling in respect to central wavelength, Doppler shift, and scanning speed to not degrade the achieved axial resolution by undersampling. For high-speed ultrahigh-resolution OCT, data acquisition boards with 16-bit or at least 12-bit resolution as well as 5 Megasamples/second or even 10 Megasamples/second or more are therefore necessary.

Finally, the mechanical performance of the scanners used for transverse and depth scanning has to be accurately selected and correctly controlled. Mechanical jitter or displacement of adjacent depth scans as well as noisy control signals of the scanner might result in distorted and therefore resolution-degraded ultrahigh-resolution OCT tomograms.

4.1.

Ultrabroad-Bandwidth Light Sources in the 800-nm Wavelength Region

The first axial resolution lower than 10 μm was achieved by using broadband fluorescence from organic dye¹²⁰ and from a Ti:sapphire laser.¹²¹ However, biological imaging could not be performed with these light sources because of their low brightness. Recently a superluminescent Ti:sapphire crystal was demonstrated as a simple and robust alternative to achieve a sub-2-μm axial resolution in OCT.¹²² The drawback of this light source is that it has to be pumped by an expensive 5-W frequency-doubled, diode-pumped laser to achieve only 40 μW of output power. Hence this light source is still neither a low-cost alternative nor suitable for in vivo clinical high-speed, ultrahigh-resolution OCT imaging. Besides the standard specifications of about 20 to 30 nm bandwidth FWHM centered in the 800- to 850-nm wavelength region, providing about 10 to 15 μm of axial OCT resolution, superluminescent diodes with up to 70-nm bandwidths (FWHM), typically centered at 830 nm with about 3-mW fiber pigtailed output power are also commercially available.¹²³ The spectrum of these SLDs is modulated in a double hump shape, which is also forward-current dependent. By multiplexing spectrally displaced superluminescent diodes to increase the optical bandwidth,^{124 125 126} improved-resolution OCT tomograms with about a 7-μm resolution in the retina have been demonstrated.¹²⁷ Recently reported cost-effective approaches for broad-bandwidth light sources mainly take advantage of the lower power demand with ultrahigh-resolution OCT imaging.^{128 129 130 131} The limiting factors of these systems are the relatively small bandwidths for ultralow-pump-threshold Kerr-lens modelocked (KLM) Ti:sapphire lasers and the strongly modulated spectra of Cr ³⁺ -ion lasers; thus they are not perfectly suitable for OCT.

Today, ultrashort light pulses produce the shortest, controlled, technically produced events. This can be demonstrated by the fact that a single femtosecond compared to a second is equivalent to 5 min compared to the age of the universe. Femtosecond laser development has concentrated so far mainly on the temporal features of the pulses, which were often optimized to the detriment of the spectral shape. In OCT, the pulse duration is irrelevant, while the spectral width and shape play a crucial role. However, unlike ultrafast femtosecond time-resolved measurements where special care must be exercised to maintain the short pulse duration, OCT measurements depend on field correlations rather than intensity correlations. Field correlation is preserved even if the pulse duration is long. Femtosecond mode-locked solid-state lasers can generate ultrabroad-bandwidth, low-coherence light with a single spatial mode and high power, providing both high resolution and the high power necessary for high-speed OCT imaging. These lasers can operate over a broad range of wavelengths that are desirable for ultrahigh resolution as well as spectroscopic OCT imaging in tissue. In a first approach, a Ti:sapphire laser has been used for in vitro OCT imaging in nontransparent tissues with a 4-μm axial resolution.¹³²

In preliminary studies, an OCT system was developed and could be optimized to support 260 nm of optical bandwidth from a state-of-the art Ti:sapphire laser.¹³³ This laser was developed in collaboration with other investigators at Massachusetts Institute of Technology and generated pulses of <5.5 fs duration, corresponding to bandwidths of more than 350 nm at an 800-nm center wavelength.¹³⁴ This high performance was achieved using specially designed double-chirped mirrors with a high reflectivity bandwidth and controlled dispersion response. Figure 8 shows a comparison of the spectra and resolution of an OCT A-scan using a conventional superluminescent diode light source versus the femtosecond Ti:sapphire laser source. The ultrabroad bandwidths generated by the femtosecond laser allow the axial resolution of OCT to be improved by a factor of nearly 10 over standard OCT technology. This femtosecond laser source has been used for imaging studies using an OCT microscope as well as an ophthalmic system interfaced to a biomicroscope system.

Figure 8

Solid-state laser light sources make ultrahigh resolution and spectroscopic OCT imaging possible. (a) The spectrum of the Ti:sapphire laser versus a standard superluminescent diode (SLD) is shown, depicting their respective wavelength bandwidths. (b) A demodulated OCT axial scan showing the axial resolution of OCT using Ti:sapphire (solid) versus SLD (dashed) light sources. Solid-state lasers provide almost a 10 times improvement in resolution.

Figure 9 demonstrates the feasibility of this novel OCT system for three-dimensional in vivo subcellular imaging of a Xenopus laevis (African tadpole) mesenchymal cell. A total of eighteen tomograms spaced 2 μm apart of an area 70×50 μm, covering approximately half of the cell, have been imaged at 1×3 μm (length×width) resolution, consisting of 170×100 pixels and 0.4×0.5-μm pixel spacing. The back surface of the cell (at 0 μm), cell membranes, and nuclei as well as intracellular morphology (at 20 to 26 μm) can be visualized. Preliminary results of an in vivo optical biopsy using ultrahigh-resolution retinal OCT imaging in human subjects have demonstrated for the first time the visualization and quantification of intraretinal structures.^{135 136}

Figure 9

Eighteen in vivo ultrahigh-resolution (1×3 μm; axial×transverse) tomograms spaced 2 μm apart with 1×3-μm resolution of half of a Xenopus laevis (African tadpole) mesenchymal cell. The cell membrane and cell nuclei as well as intracellular morphology (e.g., at 20 to 26 μm) can be visualized.

These preliminary results were obtained with an ultrahigh OCT system that is based on a complex laboratory laser system that is not suitable for clinical studies. Therefore a new generation of a compact ultrahigh-resolution OCT system was developed in collaboration with the Photonics Institute of the Technical University of Vienna and FEMTOLASERS Produktions GmbH, Vienna, Austria. This system consists of a high-speed scanning unit (up to 250 Hz, 400 mm/s) integrated in a fiber optics-based Michelson interferometer employing a compact, user-friendly, state-of-the-art sub-10 fs Ti:sapphire laser [Femtosource Compact Pro, FEMTOLASERS [Fig. 10(a)], 800-nm center wavelength, up to 165-nm (FWHM) optical bandwidth [Fig. 10(b)], 400-mW output power. The interferometer was interfaced with a microscope delivery system. Both the fiber optic interferometer and the optical components of the microscope were designed to support the propagation of very broad bandwidth light throughout the OCT system and to compensate for any polarization and dispersion mismatch between the sample and reference arms of the interferometer.

Figure 10

(a) Schematic diagram of a compact, commercially available Ti:sapphire laser (Fem Tosource Compact Pro, FEMTOLASERS GmbH). (b) Typical optical output power spectra of this Ti:sapphire laser and (c) corresponding interference signal that makes possible a free space axial resolution of 2 μm, corresponding to 1.4 μm in biological tissue.

In order to achieve high transverse resolution, a specially designed achromatic objective with a 10-mm focal length and a numerical aperture of 0.25 was used to achieve 3-μm free space transverse resolution. To overcome the depth-of-field limitation and to maintain high transverse resolution at varying depths through the image, a zone focus and image fusion technique were used. Applying laser light centered at 800 nm with up to a 165-nm bandwidth (FWHM), an axial resolution of 2.0 μm in air, corresponding to 1.4 μm in biological tissue, has been achieved with this system [Fig. 10(c)]. A signal-to-noise ratio of 105 dB was achieved at a 1-MHz carrier frequency by using an incident power of 5 mW employing dual-balanced detection. Full interference fringe signal OCT data were digitized with a high-speed (10 Ms/s) and high-resolution (16-bit) analog-to-digital (A/D) converter following software demodulation.

As a first step, ex vivo pig and monkey (Macaca fascicularis) retinal specimens were acquired to correlate ultrahigh-resolution OCT images with histology and to provide a basis for improved interpretation of in vivo ophthalmic OCT tomograms of high clinical relevance.¹³⁷ Pig retinas have an anatomy similar to that of humans; monkey retinas are identical to human retinas. For exact interpretation of OCT tomograms and correlation to histology, it is important to use fresh retinal specimens. Pig retinas were therefore chosen because of their easier availability within 1 to 2 h postmortem compared with human retina samples. Pig retinal samples were immediately preserved in buffered solutions to maintain osmolarity and oxygen support; monkey retinal samples were perfusion fixed. Figure 11 shows an example of an ultrahigh-resolution OCT image of a pig retina 2 h postmortem acquired with 1.4-μm axial and 3-μm transverse resolution. The image dimensions are 0.7×2 mm, with 20,000×2000 pixels. Enlargements show the clear visualization of nerve fiber bundles, Mu¨ller cell arcades, and all major intraretinal layers. Furthermore, the high axial resolution allows clear identification of areas of detached and nondetached retina to the choriocapillaris, which is extremely important for ophthalmic diagnosis.

Figure 11

Ultrahigh-resolution OCT image (1×3 μm; axial×transverse resolution) of a pig retina 2 h postmortem. The enlargements indicate (a) and (b) nerve fiber layer regions, (d) intraretinal layers, (e) retinal detachment, and (f) OCT scan location.

Figure 12 shows a comparison of in vitro ultrahigh-resolution OCT imaging (a) with histology (b) using a differential interference contrast (DIC) micrograph of a frozen section obtained from the matching retinal position in the pig retina. From the proximal (top) to the distal (bottom) retina, alternate dark and light bands of signal in the OCT image directly correlate with the retinal layers, nerve fiber layer (NFL), ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer (INL), and outer plexiform layer (OPL). Distal to a band attributable to the outer nuclear layer (ONL), a more delicate bright layer is likely to represent both the external limiting membrane and the myoid portion of the cone inner segments (IS). The adjacent dark and stippled signal is in alignment with the cone ellipsoids prominently seen using DIC optics. A distal dark band is possibly associated with the cone outer segment tips and, finally, the dark and light banding is attributable to the pigment epithelium–choriocapillaris complex.

Figure 12

Comparison of (a) in vitro ultrahigh-resolution OCT imaging (1×3 μm; axial×transverse) with histology and (b) differential interference contrast (DIC) micrograph of a frozen section obtained from the matching retinal position in the pig retina. From the proximal (top) to the distal (bottom) retina, alternate dark and light bands of signal in the OCT image directly correlate with the retinal layers, nerve fiber layer (NFL), ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer (INL), and outer plexiform layer (OPL). Distal to a band attributable to the outer nuclear layer (ONL), a more delicate bright ribbon is likely to represent both the external limiting membrane and the myoid portion of the cone inner segments (IS). The adjacent dark and stippled signal is in alignment with the cone ellipsoids prominently seen using DIC optics. A distal dark band is possibly associated with the cone outer segment tips and, finally, the dark and light banding is attributable to the pigment epithelium–choriocapillaris complex. Scale bar-100 μm.

Figure 13 shows a comparison of in vitro ultrahigh-resolution OCT imaging [Fig. 13(a)] with the histology [Fig. 13(b)] of a papillomacular scan of a monkey retina, demonstrating excellent correlation. Figure 14 demonstrates the ability of ultrahigh-resolution OCT to visualize specific morphological features such as the epithelium, Bowman’s layer, and Descement’s membrane and endothelium in addition to the typical stromal structure in a human cornea 4 days postmortem.

Figure 13

Comparison of (a) in vitro ultrahigh-resolution OCT imaging (1×3 μm; axial×transverse) with (b) histology of a papillomacular section of a perfusion-fixed monkey (Macaca fascicularis) retina.

Figure 14

In vitro ultrahigh-resolution OCT imaging (1×3 μm; axial×transverse) of a 4-day postmortem human cornea, visualizing all major intracorneal layers.

Recently, a compact, clinically viable ultrahigh-resolution (∼3 μm) ophthalmic OCT system has been developed and used in clinical imaging for the first time.¹³⁸ In this study, a compact, robust, commercially available Ti:sapphire laser (Compact Pro, FEMTOLASERS) with up to a 165-nm bandwidth at an 800-nm center wavelength was used. This ultrahigh-resolution OCT system is based on a commercially available OCT system (OCT 1) that was provided by Carl Zeiss Meditec, Inc. OCT imaging was performed with axial scan rates up to 250 Hz using up to 800 μW of incident power in the scanning OCT beam, which is well below the ANSI exposure limits. To date, more than 250 eyes of 160 patients with different macular diseases, including macular hole, macular edema, age-related macular degeneration, central serous chorioretinopathy, epiretinal membranes, detachment of pigment epithelium and sensory retina, glaucoma, and different hereditary retinal diseases have been examined. Figure 15 shows the appearance of the photoreceptor (PR) and outer nuclear layer (ONL) as well as external limiting membrane (ELM) in a normal human retina compared with different macular pathologies. In the extreme case of Stargadt dystrophy, the PR, the ONL, and the ELM appear completely degenerated, whereas in different stages of macular hole the PR layer as well as the ELM seem to be less affected in early stages (cf. macular hole 1 with later stages, macular holes 2 and 3). A displacement of the PR layer as well as the ELM can be seen in acute CSC and partially in chronic CSC, whereas in AMD with minimal CNV, the PR layer and the ELM are affected by CNV in the center as well as the retinal and the retinal pigment epithelium detachment. In patients with AMD and vitelliform lesion, despite retinal detachment, the PR layer, the ELM, and the ONL appear quite normal, which explains the preservation of visual acuity in these patients.

Figure 15

In vivo ultrahigh-resolution OCT images of the foveal region of a normal human retina and different macular pathologies. The arrows indicate different appearances of the photoreceptor and outer nuclear layers, the asterisks indicate impairment of the external limiting membrane. ILM, internal limiting membrane; NFL, nerve fiber layer; GCL, ganglion cell layer; IPL, OPL, inner and outer plexiform layer; INL, ONL, inner and outer nuclear layer; HF, Henle’s fiber layer; ELM, external limiting membrane; IS PR, inner segment of photoreceptor layer; OS PR, outer segment of photoreceptor layer; RPE, retinal pigment epithelium.

Figure 16(a) depicts an ultrahigh-resolution OCT image of a healthy subject. Figure 16(b) shows the same foveal region of a 74-year-old patient with endstage glaucoma, documented by visual field [Fig. 16(c)] and a Heidelberg retinal tomogram image [Fig. 16(d)]. Compared with a scan acquired with ultrahigh resolution in a healthy subject, several morphological differences are obvious. Owing to increased intraocular pressure, the foveal depression is less pronounced in the endstage glaucoma patient (light gray arrows). Nasally there is no longer any thickness to the nerve fiber layer in this patient (white thick arrows). Finally, the ganglion cell layer is atrophied (circle with white thin arrows) and has a mean thickness of about 45 μm (together with the inner plexiform layer) in the parafoveal region, compared with about 100 μm in the normal subject. In addition to the distribution of the nerve fiber layer thickness around the optic disk, this atrophy of the ganglion cell layer in the foveal region might allow a novel approach for a more sensitive and specific early diagnosis of glaucoma.

Figure 16

In vivo ultrahigh-resolution OCT detecting and quantifying subtle changes for early diagnosis in (b) endstage glaucoma patient documented by visual field test (c) and (d) Heidelberg retinal tomogram (HRT) image compared with normal subjects (a). The gray arrows indicate the foveal pit contour and the white arrows indicate the lost nerve fiber layer. Atrophy in the ganglion cell and inner plexiform layer is shown by the circle and thin white arrows.

Figure 17 demonstrates the potential of ophthalmic ultrahigh-resolution OCT to monitor and therefore contribute to a better understanding of novel therapeutic approaches. In this case the effect of an anti-vascular endothelium growth factor (VEGF) drug was investigated. This case was a 60-year-old woman with regressing drusen and placoid occult without classic choroidal neovascularization [Figs. 17(a) to 17(c)]. In addition to a normal appearance of the NFL, GCL, IPL, and INL, the photoreceptor layer is strongly compromised, owing to a serous detachment. Ultrahigh-resolution OCT makes it possible to visualize the occult membrane underneath the RPE as well as the outer layer of Bruch’s membrane, indicating intra-Bruch’s choroidal neovascularization for the first time (arrows). In all three cross-sections, highly reflective particles in the outer photoreceptor layer, which could be isolated RP cells or damaged photoreceptors, could be differentiated by ultrahigh-resolution OCT. Fifteen weeks later, the effect of therapy treatment can clearly be visualized with ultrahigh-resolution ophthalmic OCT [Fig. 17(d) to 17(f)]. The PR complex has recovered but some subtle serous detachments as well as the occult membrane are still present. Finally, Fig. 18 demonstrates the potential of ophthalmic ultrahigh-resolution OCT to not only provide additional information about the status of the photoreceptor layer in patients with macular hole that can be used in a decision on surgical intervention, but it also may allow monitoring of the effect of this intervention and investigation of its outcome.

Figure 17

Monitoring of new therapy approaches using in vivo ultrahigh-resolution OCT. A patient with regressing drusen with placoid occult without classic choroidal neovascularization before (a)–(c) and 15 weeks after treatment with anti-vascular endothelium growth factor (d)–(f).

Figure 18

Monitoring of surgical intervention with in vivo ultrahigh-resolution OCT. A patient with a macular hole is shown before (a) and after (b) surgery.

These preliminary results demonstrate that ultrahigh-resolution ophthalmic OCT enables unprecedented visualization of intraretinal morphology, which previously had been possible only with histopathology. It therefore provides a powerful tool for the clinical diagnosis of retinal diseases that are the leading causes of blindness. Furthermore, ultrahigh-resolution OCT provides a basic research and clinical tool for investigating the impairment of all major intraretinal layers, especially the photoreceptor layer associated with different macular pathologies. Its increased accuracy may allow the detection of intraretinal changes that can be used to diagnose retinal disease in its early stages, when treatment is most effective and irreversible damage can be prevented or delayed. Furthermore, it may provide a better understanding of the pathogenesis of several macular pathologies as well as the development of new therapeutic approaches.

The ultimate availability of this new technology for clinical research and patient care depends mainly on the availability of compact, reliable, low-cost lasers or other sources for ultrabroad-bandwidth light. Recently a compact, prismless, low-cost Ti:sapphire laser that employs a commercially available pump source has been used for in vivo ultrahigh-resolution imaging of macular pathologies.¹³⁹ It contains a standard astigmatism-compensated X-folded cavity with a compact low-cost 1.5-W pump source (158×104 mm) [Fig. 19(a)]. Dispersive mirrors are used to compensate for the second- and third-order dispersion of the laser crystal and air. Special broadband, low-loss dispersive mirrors were designed. The resonator is asymmetrically folded and some mirrors are used in double pass to reduce the laser size. The advantage of this prismless laser is that the design of the cavity is more compact and can act as a hands-off OCT laser source. The overall dimensions of the setup are 500×200 mm, including the pump laser. Kerr-lens mode locking is initiated by rapid translation of the end mirror, which is mounted on a translation stage to induce intensity fluctuations. The output power of the laser was strongly dependent on the bandwidth of the laser and varied from 100 mW with 135 nm to 20 mW with a 176-nm FWHM and a 70-MHz repetition rate [Fig. 19(b)]. For ex vivo OCT imaging, the broadest possible bandwidth [176 nm centered at 776 nm, see Fig. 19(b), black line] provided a peak axial resolution of 1.7 μm in free space, corresponding to about 1.2 μm in tissue [Fig. 19(c)].

Figure 19

Setup of the compact, low-cost Ti:sapphire laser. A commercially available compact (158×104 mm) 1.5-W source is used to pump a standard astigmatism-compensated X-folded cavity. (b) Typical optical output power spectra of this laser. (c) Interference signal (corresponding to the spectrum indicated with the solid black line, resulting in a free space axial resolution of 1.7 μm, corresponding to 1.2 μm in biological tissue).

This laser was interfaced to a fiber-based OCT system and achieved the same performance for axial resolution, sensitivity, and stability as a Ti:sapphire laser pumped by a 5-W single-frequency, frequency-doubled, diode-pumped neodymium:vanadate laser.¹³⁹ For ultrahigh-resolution in vivo imaging in normal persons and patients, a less-modulated, Gaussianlike spectrum was used, compromising the bandwidth to 135 nm FWHM at 95 mW [Fig. 19(b), gray line] and obtaining an axial resolution in the retina of about 3 μm, which is similar to recent results. The system exhibits extremely reproducible spectra, output power, and user friendliness in day-to-day performance. Stable operation for more than 12 h with a 2 power loss and 2 to 5 loss in spectral bandwidth are usual.

An even further increase in the bandwidth using the same oscillator design can be achieved by replacing all high-reflecting mirrors by broadband dispersive mirrors and employing a single-frequency 532-nm pump source, which increases the footprint to 600×400 mm. The spectrum generated was 265 nm FWHM (Fig. 20). The bandwidth was only limited by the dichroic input coupler with an overall bandwidth of 300 nm. The average mode-locked output power is 250 mW with 3.65 W of pump power. This laser therefore has large potential to be used in combination with a low-cost pump laser and may be extremely interesting not only for in vivo submicrometer-resolution OCT, but also for functional, spectroscopic OCT.

Figure 20

(a) Setup of the compact, ultrabroad-bandwidth Ti:sapphire laser. (b) Typical optical output power spectra of this laser.

4.2.

Ultrabroad-Bandwidth Light Sources in the 1300-nm Wavelength Region

For OCT to be successfully used as an in vivo optical biopsy tool in nontransparent tissue, micrometer-scale resolution and millimeter-range penetration depth are required. In the range between 0.8 and 1.8 μm, scattering is the predominant mechanism limiting image penetration depth. Since scattering depends strongly on wavelength and decreases for longer wavelengths, a much higher image penetration depth in nontransparent tissue can be achieved with light at 1.35 μm than at 0.85 μm. It has been shown that the optimum wavelengths for imaging in nontransparent biological tissues are in the 1.3- to 1.5-μm range.^{115 140} In this range, imaging depths of 2 to 3 mm can be achieved. Narrow-bandwidth superluminescent diodes or multiquantum well semiconductor amplifier used for OCT so far typically provide axial resolution of 10 to 20 μm. By multiplexing spectrally displaced light-emitting diodes (LEDs) at 1300 nm, a 7.2-μm axial resolution in free space could be obtained.⁹⁷

In the 1300-nm wavelength region, ultrashort-pulse solid-state lasers are also promising light sources for ultrahigh-resolution OCT. In a first approach, a self-phase-modulated KLM chromium:forsterite (Cr:forsterite) laser has been used for in vivo OCT imaging in nontransparent tissues with 6-μm axial resolution.^{74 141} Recent efforts were focused on developing even broader bandwidth light sources in the 1300-nm wavelength range that would permit OCT micrometer-scale resolution, along with millimeter-range penetration depth. A laser spectrum covering the 1230- to 1580-nm wavelength region with an optical bandwidth of 250 nm (FWHM) was generated directly out of an all-solid state Cr:forsterite laser.¹⁴²

Spectra far broader than one optical octave can be produced via nonlinear propagation of laser pulses in microstructured fibers. Owing to the geometry of these fibers, the cross-section of the fundamental mode is unusually small, which enhances the peak power and thus the nonlinearity. At the same time, the fiber dispersion can be engineered to avoid fast temporal spreading. Owing to these effects, spectra covering more than one optical octave can be produced, even with pulses having moderated energies of a few nanojoules. This spectral width could never be generated directly from the laser oscillator because it exceeds the fluorescence bandwidth of the crystal. In order to avoid an excessively strong spectral modulation (inherent to the spectral broadening processes—self-phase modulation and four-wave mixing), only moderate spectral broadening should result from the nonlinear fiber propagation, i.e., the initial bandwidth of the pulses emerging from the oscillator should be as broad as possible. High nonlinearity, air-silica microstructure fibers,¹⁴³ or tapered fibers¹⁴⁴ have been used to generate an extremely broadband continuum using low-energy femtosecond pulses. In the latter approach, a standard single-mode fiber is heated and then lengthened. The light is initially guided inside the fiber core, but in the tapered region, the light propagates through the cladding. As in the air-hole fibers, the change in the index of refraction between the light-guiding region, the cladding in the tapered fibers, and the surrounding region, the air, is high. Depending on the cladding diameter, the zero-dispersion wavelength can be shifted. Both the air-hole fiber and the tapered fiber have been studied to find out optimum conditions for generating a broadband continuum with minimum noise.

So far, the highest reported OCT axial resolution in biological tissue in the 1300-nm wavelength region was obtained with the use of a complex laboratory prototype photoionic crystal fiber (PCF)-based light source.¹⁴⁵ With a bandwidth of 450 nm at a 1.3-μm center wavelength [Figs. 21(a) and 21(b)], a 2.5-μm axial resolution in free space [Fig. 21(c)], which corresponds to 2 μm in tissue, was achieved by employing an OCT system that could support a 370-nm bandwidth [Fig. 21(c)]. Figure 21(d) depicts an in vivo ultrahigh-resolution image of a Syrian hamster’s cheek pouch and Fig. 21(e) shows the corresponding histology. Recently, an unprecedented axial resolution of less than 2 μm in the 1300-nm wavelength range in nontransparent biological tissue was achieved with a novel, broad-bandwidth light source (MenloSystems), which due to its compactness, low cost, stability, and user friendliness offers a great potential for OCT clinical applications.¹⁴⁶ The MenloSystems light source is based on a pulsed erbium fiber laser, the amplified output of which is coupled into a highly nonlinear fiber to generate a supercontinuum ranging from 1100 to 1800 nm at its pedestal, with a power output of about 50 mW at a 50-MHz repetition rate. The light source is very compact (145×480×300 mm, corresponding to a 19-in. rack, 3 height units), low weight (8 kg), and has high mechanical stability. It provides a single mode fiber (SMF) 28 fiber-coupled output, which simplifies the interface to fiber optic OCT systems. Because of a 12-dB modulation at about 1550 nm, only the shorter wavelength portion of the emission spectrum was utilized for OCT imaging by blocking the wavelength range beyond 1550 nm with an edge filter. The filtered fiber laser spectrum was moderately modulated, centered at 1375 nm with a full width at half maximum of 470 nm and a power output of 4 mW (Fig. 22). By interfacing the MenloSystems light source to a free-space OCT system, a resolution of 2 μm in an axial, and 4 μm in a lateral direction (measured with a resolution target), corresponding to ∼1.4 μm and ∼3 μm in biological tissue, respectively (Fig. 22), and 95-dB sensitivity for 500-μW incident power were achieved.

Figure 21

Ti:sapphire-pumped PCF-based light source. With a bandwidth of 450 nm at a 1.3-μm center wavelength (a), 2.5 μm of axial resolution in free space (c), which corresponds to 2 μm in tissue, was achieved by employing an OCT system that could support a 370-nm bandwidth (b). In vivo ultrahigh-resolution imaging of a Syrian hamster’s cheek pouch could be achieved. (d) Corresponding histology. (e) From Ref. 145 (with permission of Prof. J. G. Fujimoto, Cambridge, MA, USA).

Figure 22

(a) Typical optical output power spectra of a compact, ultrabroad-bandwidth fiber laser and (b) an interference signal corresponding to the spectrum that produces a free space axial resolution of 2 μm, corresponding to 1.4 μm in biological tissue. (c) to (e) Ex vivo ultrahigh-resolution OCT tomograms obtained with this light source.

The feasibility for an ex vivo sub-2-μm axial resolution OCT was demonstrated by imaging nonfixed human skin, arterial as well as gynecological biopsies (Fig. 22) demonstrating that the MenloSystems light source has great potential for applications in in vivo OCT clinical studies. Owing to its extremely broad output spectrum, this fiber laser-based light source also covers the 1550-nm wavelength region. Recent studies employing erbium- and thelium-doped fiber laser at 1.55 and 1.81 μm have demonstrated that these wavelength regions are interesting for OCT imaging of nontransparent tissue, because of good penetration depth despite the increased water absorption.¹⁴⁷ This might be due to the significantly reduced scattering in this wavelength region. Other broad-bandwidth light sources in this wavelength region that might be useful for OCT include a prismless 20-fs pulse Cr ⁴⁺: yttrium aluminum garnet (YAG) laser that offers a 190-nm bandwidth (FWHM) centered at 1400 nm with a 400-mW output power,¹⁴⁸ or an extremely powerful (2.4-W), ultrabroad-bandwidth (900-nm) compact white-light source that employs a passively mode-locked Nd: vanadate pump laser together with a dispersion-adapted air-silica microstructured fiber.¹⁴⁹

4.3.

New Wavelength Regions for Ultrahigh Resolution OCT

In the near-infrared region, a novel OCT light source might offer a unique opportunity to perform ophthalmic OCT imaging at 1100 nm, a spectral region where the transparency of ocular media is still good but the penetration below the retinal pigment epithelium could be significantly enhanced. This might be very important not only for visualizing the choriocapillaris and choroid in the detection of diabetic macular edema but also for the detection of the angiogenesis process in age-related macular degeneration. Moreover, longer wavelengths should cause less attenuation in opaque eye media, which might occur in older patients as a result of cataract and haze in the cornea. Therefore, a photonic crystal fiber-based light source centered at about 1050 nm has recently been developed and was used to evaluate the image penetration depth of ultrahigh-resolution OCT in the choroid compared with a spectrally broadened fiber laser [λ_c=1350 nm, Δλ=470 nm, P _out =4 mW; Figs. 23(a) and 23(b)] as well as a state-of-the art Ti:Al ₂ O ₃ laser [λ_c=780 nm, Δλ=160 nm, P _out =400 mW; Figs. 23(a) and 23(b)].¹⁵⁰ A PCF with a 1.9-μm core size and a 24-mm length was pumped with the output of a self-starting Ti:Al ₂ O ₃ laser (FEMTO-LASERS; λ_c=790 nm, Δλ=33 nm, P _out =300 mW; Fig. 23) to generate a broad-bandwidth spectrum, covering a local minimum in the water absorption spectrum located at about 1060 nm [Figs. 23(a) and 23(b)]. The PCF-generated supercontinuum, spanning 400 to 1200 nm, with P _out =60 mW, was filtered with a long-pass filter to produce a spectrum ranging from 900 to 1200 nm with P _out =10 mW. A free space resolution of 3.5×4 μm (lateral×axial), corresponding to 2.5×3 μm in tissue and 98 dB sensitivity at 2-mW of incident power was achieved by interfacing the PCF-based source to the OCT system.

Figure 23

(a) Output spectra of a Ti:sapphire (thin black line left), PCF-based source (gray line), and a MenloSystems laser (black line, right), overlaid with a water absorption spectrum (dashed line). (b) Corresponding fringe patterns produced by interfacing the light sources to the OCT system. (c) to (e) Ex vivo OCT images of pig retinas acquired with (c) the Ti:sapphire source (at about 800 nm, 2000×1000 pixels, 2×1 mm), (d) a PCF-based source (at about 1050 nm, 2000×1010 pixels, 2×1 mm), and (e) the fiber laser (at about 1350 nm, 2000×888 pixels, 2×1 mm) demonstrating enhanced penetration into the choroid with increasing wavelength.

The ex vivo pig retina tomogram [Fig. 23(c) 2000×1010 pixels, 2×1 mm] acquired with the Ti: sapphire laser demonstrates that ultrahigh-resolution OCT in the 800-nm wavelength region permits detailed visualization of intraretinal layers. In comparison, the pig retina tomogram acquired with the PCF-based light source [Fig. 23(d), 2000×1010 pixels, 2×1 mm] shows better penetration in the choroid (two layers of blood vessels are observed). Superior penetration in the choroid was achieved with the MenloSystems fiber laser, as demonstrated in the retinal tomogram in Fig. 23(e) (2000×888 pixels, 2×1 mm), where a group of interlaced vessels is clearly visible. This study demonstrated that with light sources centered at about 1050 and 1350 nm, the penetration depth of the OCT image in retinal tissue at longer wavelengths was superior. Compared with images acquired at about 800 nm, better visualization of the choriocapillary and choroid was achieved with the PCF-based source (∼1050 nm). Superior penetration into the choroid was realized with the MenloSystems laser (∼1350 nm), though owing to enhanced water absorption at longer wavelengths and the axial length of a human eye (∼25 mm), this light source cannot be used for in vivo imaging of human retina. The fiber laser may find clinical applications for in vivo imaging of the anterior segment of a human eye. Furthermore, the fiber laser can be applied successfully in ophthalmic animal studies¹³ since the eye length of a mouse is about 5 to 7 mm.

Spectra with up to 372 nm centered at about 1100 nm with about 50 mW of output power were recently generated by pumping a PCF with a 700-mW, 110-fs Ti:sapphire oscillator [Figs. 24(a) and 24(b)].¹⁵¹ Nitrogen gas was used to avoid damage on the PCF fiber tip. An axial resolution of 1.8 μm was achieved [Figs. 24(c) and 24(d)] and in vivo ultrahigh-resolution OCT imaging was demonstrated on an African tadpole (Xenopus laevis) [Figs. 24(e) to 24(f)]. This light source offers the broadest bandwidth ever achieved in the 1100-nm wavelength region. The disadvantage of ultrabroad-bandwidth spectra centered at 1100 nm is the lack of highly sensitive detectors for this wavelength region, since the spectrum covers a large portion of the detector sensitivity interface between silicon and indium-germanium-arsenic (InGaAs) detectors. Germanium or Si–InGaAs sandwiched detectors might be used, but are less sensitive than silicon and InGaAs detectors alone.

Figure 24

(a) and (b) Spectra with up to 372 nm centered at about 1100 nm with about 50 mW of output power generated by pumping a PCF with a 700-mW, 110-fs Ti:sapphire oscillator achieving (c) and (d) 1.8-μm axial resolution. (e) and (f) In vivo ultrahigh-resolution OCT imaging on an African tadpole (Xenopus laevis) is depicted with (e) and without (f) dynamic focusing. (From Ref. 151.) (with the permission of Prof. Z. Chen, Beckman Laser Institute, University of California, Irvine, CA, USA)

Recently, superluminescent diodes that cover a broad bandwidth in a similar wavelength region have become commercially available. These SLDs offer up to 70 nm of bandwidth (FWHM) centered at 950 nm with typically 3-mW ex fiber pigtail power as well as more than 100 nm of bandwidth (FWHM) centered at 900 nm with 3-mW free space output power. Unfortunately the output spectra of these light sources coincide with the water absorption peak around 980 nm. Therefore a portion of the output power will be absorbed and attenuated, especially in the case of ophthalmic OCT, where the beam has to propagate through 25 mm of ocular media consisting mainly of water.

For the first time, submicrometer OCT resolution has recently been achieved by using a PCF pumped with a compact, commercially available sub-10-fs Ti:sapphire laser. A stable, slightly modulated spectrum ranging from 550 to 950 nm (at its pedestal) with an output power of few tens of milliwatts was generated with this source, thus allowing in-depth imaging of biological tissue with unprecedented axial resolution.¹⁵² The broad bandwidth of the light source also provides access to a spectral region covering the absorption bands of a number of biological chromophores; thus it has also great potential for spectroscopic OCT.

A 2.3-μm core diameter, 6-mm length PCF (University of Bath, UK) was pumped with a compact, commercially available Ti:sapphire ( Al ₂ O ₃ ) laser (Femtosource Compact Pro, FEMTOLASERS) emitting sub-10-fs pulses with about a 100-MHz repetition rate and an output power of up to 400 mW. By proper selection of the input pulse parameters and PCF characteristics, a spectrum spanning 325 nm, centered at 725 nm, with spectral modulations less than 1.5 dB and power output up to 27 mW was generated [Fig. 25(a)]. By interfacing the PCF source with a free space interferometer optimized to propagate the full bandwidth with minimum power and spectral losses and using a standard resolution chart, a free space OCT resolution of 2.5 μm in the lateral and 0.75 μm in the axial direction [corresponding to 0.5 μm in biological tissue; Fig. 25(b) has been achieved]. The axial resolution was defined as the FWHM of the envelope of the measured interferogram.

Figure 25

(a) Optical spectrum of PCF output (solid line), PCF input spectrum (dashed line), and (b) corresponding interference signal.

To demonstrate OCT in tissue with sub-micrometer axial resolution, human colorectal HT-29 adenocarcinoma cells were imaged in vitro. Multiple OCT cross-sectional images of a group of HT-29 cells were acquired in vitro with 0.5-μm axial and about 2-μm transverse resolution, covering an area of 50×50 μm (500×500 pixels) and equally spaced by 2 μm [Fig. 26(a)]. Histological sections of typical HT-29 cells parallel and perpendicular [Figs. 26(b) and 26(c)] to the OCT imaging direction are displayed for better interpretation of the OCT tomograms. A comparison with the histology shows that OCT images with a submicrometer resolution reveal features that may correspond to subcellular structures such as nucleoli that have a diameter of approximately 3 to 5 μm, as well as aggregates of cellular organelles.

Figure 26

(a) In vitro submicrometer-resolution OCT images of HT-29 human colorectal adenocarcinoma cells with 0.5-μm axial and about 2-μm transverse resolution, covering an area 50×50 μm, equally spaced by 2 μm. The arrows indicate features that may correspond to nucleoli with about a 3 to 5-μm diameter. Histological sections are shown parallel (b) and perpendicular (c) to the OCT imaging direction of typical HT-29 cells. [small black arrows in (c) indicate OCT imaging plane.]