Optical coherence tomography (OCT) is one of the well-developed biomedical imaging modalities for visualizing in vivo tissues with a depth resolution of a few micrometers.1 In OCT, the bandwidth of the probe-beam dominates its depth resolution, which is inversely proportional to the bandwidth. Light sources with bandwidths wider than are commercially available, and those with much wider bandwidths are available in laboratories. OCT systems with resolutions of few micrometers or submicrometers have been demonstrated2, 3 by taking advantage of the wide bandwidth.
Despite its excellent resolution, the long measurement time of OCT is its disadvantage. Since OCT is a point-detection system, 2-D mechanical scanning is required to obtain a 2-D tomogram, and this high-order mechanical scanning results in a relatively long measurement time. Although rapid scanning optical delay lines (RSODs) have dramatically reduced the measurement time,4, 5, 6 the A-scan rate is still around few kilohertz.7 Even if faster delay lines were devised, ensuing high-speed detection could reduce the system sensitivity, which is proportional to the light exposure time.8, 9 This trade-off between the measurement time and the sensitivity should be overcome by new modalities in order to build a high-speed and highly sensitive OCT.
Fourier domain OCT [FD-OCT or spectral domain OCT (SD-OCT)] has been studied10, 11 in recent years for this purpose. FD-OCT does not require the mechanical A-scan due to its wavelength-resolving detection system or wavelength-swept light source, and enables real-time12 and video rate cross-sectional imaging.13 Another advantage of FD-OCT over conventional time-domain OCT (TD-OCT) is its higher sensitivity. This has been proved both theoretically and experimentally.8, 14 These advantages of FD-OCT effectively address the mentioned trade-off. Furthermore, FD-OCT provides direct accessibility to the phase distributions of OCT images, thus enabling smart polarization sensitive FD-OCT (Refs. 15, 16, 17) and Doppler FD-OCT (Refs. 18, 19, 20). However, FD-OCT still requires a 1-D mechanical B-scan for acquiring a 2-D OCT image or a 2-D mechanical scan for acquiring a 3-D OCT volume.
Full-field OCT [FF-OCT or full-field optical coherence microscopy (FF-OCM); Ref. 21] is another innovative modality that successfully deal with the trade-off. With a single shot of an area detector, FF-OCT visualizes an en face image of a sample in a specified depth. This parallel detection scheme enables high resolution,22, 23 high-speed measurements,24 with high sensitivity23 because the exposure time of each pixel of the array detector can be identical to the measurement time for a single en face image. Although FF-OCT has all these advantages, it does not provide high-speed cross-sectional imaging because in FF-OCT, cross-sectional imaging is always accompanied with 3-D volume measurement by mechanical depth scanning.
In this paper, the 1-D mechanical B-scan of FD-OCT is eliminated by using25 an imaging optics as FF-OCT (Sec. 2.2). This OCT system, referred to as line-field FD-OCT (LF-FDOCT) in this paper, visualizes a cross-sectional OCT image without any mechanical scanning. A 1-D transversal mechanical scan (C-scan) is introduced into the LF-FDOCT for 3-D imaging (Sec. 2.3) and an in vivo human fingerpad is investigated in three dimensions (Sec. 3). Although 3-D LF-FDOCT has been demonstrated with static mechanical samples with relatively slow measurement speed,26 and 2-D LF-FDOCT has been applied to in vivo biological investigations,27 to the best of our knowledge, this is the first demonstration of 3-D in vivo LF-FDOCT. The advantages of the LF-FDOCT in terms of its sensitivity and measurement time over conventional scanning OCT are discussed in Sec. 4.
Figure 1 shows the optical scheme of LF-FDOCT consisting of a 2-D spectrometer, a free-space Michelson interferometer, and a cylindrical lens (CL). This scheme is nearly identical to that of conventional FD-OCT, with the exception of the cylindrical lens and the 2-D CCD. The light source is a superluminescent diode (SLD) with a center wavelength of , a bandwidth of , and maximum output power of (BWC-SLD-11, B&W TEK Inc., Delaware, USA), and it provides a depth resolution of with a refractive index of 1.38. The spectrometer consists of a VHG ( , Wasatch Photonics, Inc.), an achromatic lens (L4) with a focal length of and a diameter of (Thorlabs, Inc.), and a 2-D CCD camera (TI-124B, Nikko Denki Tsushin Co., Ltd., Japan). The diffraction angle of the grating is optimized to maximize the diffraction efficiency. The CCD camera is an NTSC-standard video camera that uses for spectral resolution and for lateral resolution (B-scan), the details of which are described later. The signals from the CCD camera are digitized into numerical values by an NTSC frame grabber (PCI-1409, National Instruments) with a detection speed of (fps). Each spectrum that is detected by the CCD camera and the frame grabber is rescaled into frequency space by spline or zero-filling interpolation,13, 28, 29 then Hanning windowed, zero-filled from , and digitally inversely Fourier transformed on a computer to obtain a depth-resolved OCT image.
The relationship between the real depth and the pixel position in the image, which depends on spectrometer specifications, is examined by a plane mirror sample mounted on a micrometer. Figure 2a shows the 1-D OCT signals of the mirror sample with different depths, and Fig. 2b shows the relationship between the real depth positions of the mirror sample and the peak positions of the OCT signals in terms of pixels. On the basis of the gradient of this plot, the calibration coefficient is determined to be . The depth measurement range is , which equals . However, half of this range is occupied by mirror images,30, 31 and the effective depth-measurement range is .
A theoretical effective depth-measurement range can be calculated from the parameters of the spectrometer byis the pixel spacing of the CCD in wavelength. In our LF-FDOCT, the pixel spacing of yields the effective measurement range of , and it agrees well with the experiment.
The LF-FDOCT does not require a mechanical A-scan due to its FD-OCT configuration. Moreover, the B-scan is not required because of line illumination and a 1-D imaging optics. As shown in Fig. 1, a CL ( focal length, Sigma Koki Co., Ltd., Japan) is placed between a collimator lens [L1; a microobjective of 0.13 numerical aperture (NA), Olympus, Japan] and a beamsplitter (BS). As illustrated in Fig. 3a , the cylindrical lens does not affect the horizontal optical perspective, and the optics is identical to that of conventional FD-OCT. On the other hand, in the vertical optical perspective, the cylindrical lens extends the illumination on the sample into a 1-D line, as shown in Fig. 3b. The sample is then illuminated by the line focus, and the illuminated area is imaged on the CCD camera by objective L2 (an achromatic lens with a focal length of and a diameter of , Thorlabs, Inc.) and a spectrometer-lens L4 with a transverse magnification of 2.5.
The vertical optical resolution of the LF-FDOCT system is defined by the NA of L2 by using Rayleigh’s criterion:is the NA of L2. Based on the already mentioned parameters, the vertical optical resolution becomes . Although this system has an optical resolution of , the resolution is also restricted by the pixel size of the CCD camera. A single pixel of the CCD corresponds to an area with a length of on the sample, hence the practical resolution is dominated by the CCD pixel size. Because of Nyquist sampling theorem, the resolution can be defined as twice the size of the CCD pixel. In our LF-FDOCT, .
The vertical measurement range of the LF-FDOCT system is dominated by the magnification and the size of the CCD, or the height of the line focus. In the proposed LF-FDOCT, the first and the second parameters dominate the measurement range of , and this field is sampled by on the CCD.
Configuration for 3-D Measurement
The LF-FDOCT can measure a 2-D tomogram without mechanical scanning. To extend LF-FDOCT for 3-D measurements, we introduce a 1-D mechanical C-scan. A gold-protected mirror mounted on a 1-D galvanometer (Model 6220, Cambridge Technology) is placed at the back focal plane of L2, and it swings the probe beam horizontally. A function generator board with a digital-to-analog resolution (DAQ-2502, ADLINK Technology, Inc.) plugged into a Windows 2000 computer generates the control signal for the galvanometer. The function generator board, and thus the galvanometer, are synchronized with the operation of the CCD camera by a trigger signal from the frame grabber. Owing to this synchronization scheme, the 3-D measurement speed of this LF-FDOCT system is limited only by the driving speed of the CCD. The B-scan rate of the 3-D acquisition is 30 B-scans∕s and it is identical with that of 2-D acquisition.
The maximum horizontal measurement field— —is dominated by the maximum rotation angle of the galvanometer and the focal length and the diameter of L2.
The horizontal resolution is defined in the same manner as that in conventional TD-OCT and FD-OCT, i.e., it is based on probe beam diameter and the focal length of the objective. In the proposed LF-FDOCT, the horizontal resolution is .
We investigate an in vivo fingerpad of a healthy human volunteer (age 24, male) in three dimensions. The probe power is sufficiently weaker than the ANSI safety standard.32 To reduce the motion artifacts, the exposure time is set to by the electrical shutter of the CCD camera (see Sec. 4.2), and 300 B-scans are acquired in . In this investigation, the separation between adjacent B-scans is and the horizontal field for the 300 B-scans is . The measurement volume size is 2.1 (vertical) 1.4 (horizontal) 1.3 mm (depth), and the corresponding data points are 480 points (vertical) 300 points (horizontal) 320 points (depth).
Figure 4 shows vertical cross sections of the sample at several horizontal positions, and a single cross section is acquired from a single-shot image on the CCD. The exponential attenuation along the depth penetration is numerically compensated. The acquisition speed is 30 cross sections∕s. Structures such as stratum corneum (SC), epithelium (EP), and dermis (DM) can be seen. The direction of the sectioning is parallel to the grooves of the fingerprint, thus, some sweat ducts (SDs) are aligned in a single B-scan.
Horizontal cross sections reconstructed from the 3-D volume at several vertical positions are shown in Fig. 5 . In this case, the direction of the sectioning is perpendicular to the fingerprint, and the grooves of the fingerprint can be seen. A white arrow points the cross section of a sweat duct which screwing up from the dermis to the stratum corneum along a virtual horizontal scanning.
En face images reconstructed from the 3-D OCT volume are shown in Fig. 6 . The ridges of the stratum corneum associated with the fingerprint are shown in the en face images corresponding to the depths of . The small aligned spots in the image corresponding to the depth of indicate SDs. The images corresponding to the depths of and show the ridges between the EP and DM.
We examine the sensitivity of the LF-FDOCT system by measuring a mirror sample with an exposure time of , a probe power of , and a probe attenuation of . The dynamic range is , therefore the system sensitivity is , while the theoretical sensitivity is . The degradation in the sensitivity may be due to the nonuniformity of the line illumination and vignetting of the CCD. Because an SLD is a spatially coherent light source, coherent crosstalk among the lateral points, which is not evident in the measurement of the mirror but is evident in that of biological samples, may cause additional degradation of the sensitivity. This coherent crosstalk can be suppressed by using spatially incoherent light source with a line slit, which is used to enhance the spatial coherence in one dimension, although this configuration sacrifices great portion of the optical power of the light source.
The exposure time dependency of the signal-to-noise ratio (SNR) and the sensitivity are examined, as shown in Fig. 8 . Figure 8a reveals that the SNR is proportional to the exposure time, as theoretically predicted, and Fig. 8b shows the theoretical shot-noise-limited sensitivity (solid curve) and experimentally measured sensitivity (×). A formula from Ref. 9 and the following parameters are used to calculate the theoretical sensitivity; , 6.5% quantum efficiency of the camera, and 80% spectrometer efficiency. The fill factor of the CCD camera, which is not disclosed by its manufacturer, is assumed to be 100%. This assumption may have resulted in the disagreement between the theoretical curve and the experimental values.
As shown in Fig. 8, a longer exposure time leads to higher sensitivity, and we can use a longer exposure time in LF-FDOCT than in conventional FD-OCT. This is because in an LF-FDOCT, the exposure time can be identical to the acquisition time of a single B-scan while the maximum exposure time of a conventional FD-OCT is equal to the acquisition time of a single A-scan. Hence, in principle, the sensitivity of LF-FDOCT is greater than that of a conventional FD-OCT by a factor of , where is the number of A-scans in a single B-scan. This discussion is analogous23 to that of an FF-OCT. In our setup, the maximum gain in the sensitivity is . However, note that the long exposure time enhances the motion artifacts and may degrade the sensitivity. The effect of the motion artifacts is discussed in the following section.
The system sensitivity also depends on the depth in the measurement field, as seen in Fig. 2a. This figure indicates that the sensitivity at the end of the measurement range is lower than the highest sensitivity, which is at the center of the measurement range (zero delay point). This degradation may be due to the limited resolution of the spectrometer, because the dimensions of optical elements limit the resolution so that it is slightly worse than the sufficient resolution. And the numerical error of the wavelength-frequency rescaling also may be the cause of the degradation.
Effect of Motion Artifacts
Although a longer exposure time leads to the higher sensitivity, it can deteriorate the image quality because of the motion artifacts.34
The effect of the motion artifacts in the LF-FDOCT is identical to that in a conventional FD-OCT, however, this effect is relatively large in the former because the exposure time is longer than that in a conventional FD-OCT. On the basis of a formula derived by Yun, 34 the maximum allowable velocity for the SNR decay of along the depth isis the exposure time, which is in our case. This result suggests that the maximum displacement in the exposure time is .
The maximum allowable lateral velocities are also calculated from the formula by Yun as for vertical motion and for horizontal motion, where and are vertical and horizontal resolutions, respectively.
These estimations suggest that the LF-FDOCT is approximately 20 times more sensitive to axial sample motion than to transversal motions.
Potential System Speed
In the measurement described in Sec. 3, the exposure time for a single B-scan is . Within this duration, 480 A-scans are acquired, hence, the A-scan rate of a single B-scan is , however, the overall A-scan rate for the 3-D measurement is because of the limited duty cyle of 3%, which is due to the fixed frame rate of the NTSC camera. However, if we use a high-speed CCD camera and a high-speed sampling device, the overall A-scan rate for 3-D measurements can be improved up to . This would potentially enable a 2-D OCT measurement at a rate of , and a 3-D OCT measurement at a rate of .
With the exception of the cylindrical lens, all of the lenses in the LF-FDOCT system are achromatic doublets, which are designed to correct both achromatic and monochromatic aberrations. The cylindrical lens, which is a planoconvex singlet, can cause an achromatic aberration. The spectral width at the top and bottom perimeters of the line focus can be slightly degraded by the aberration when compared to the spectral width of the light source itself. However, these perimeters are not used for the OCT measurement, therefore, the effect of the achromatic aberration is negligible.
The cylindrical lens also causes a comatic aberration. If the rays do not impinge perpendicularly on the cylindrical lens, the comatic aberration can degrade the quality of the line focus. However, the proposed LF-FDOCT is free from comatic aberration, because the rays enter the cylindrical lens perpendicularly, as shown in Fig. 3a.
We demonstrated a new modality of 3-D OCT. This system eliminates its A-scan by using FD-OCT and avoids the mechanical B-scan by an imaging optics. The system has a depth resolution of , a vertical lateral resolution of , and a horizontal lateral resolution of . The system sensitivity is experimentally determined to be with a probe power of and an exposure time of . The A-scan rate of the system is for a single B-scan, while it is for a 3-D measurement.
The system visualized inner structures of an in vivo human fingerpad in an area of 2.1 (vertical) 1.4 (horizontal) 1.3 mm (depth) for . The corresponding image size is 480 points (vertical) 300 points (horizontal) 320 points (depth), while the maximum measurement size is 480 points (vertical) 300 points (horizontal) 1024 points (depth). In this measurement, the maximum allowable velocities are for axial sample motion, for vertical motion, and for horizontal motion.
This research is partially supported by Grant-in-aid for Scientific Research No. 15760026 from the Japan Society for the Promotion of Science (JSPS), Japan Science and Technology Agency, and the Special Research Project of Nanoscience at University of Tsukuba. Shuichi Makita is supported by JSPS by a contract under the Promotion of Creative Interdisciplinary Materials Science for Novel Functions, 21st Century Center of Excellence (COE) Program. The authors wish to acknowledge Dr. Masahiro Akiba of the Yamagata Promotional Organization for Industrial Technology for helpful discussions.