We demonstrate a compact, ultrahigh speed spectral-domain optical coherence microscopy (SD-OCM) system for multiscale imaging of specimens at 840 nm. Using a high speed 512-pixel line scan camera, an imaging speed of 210,000 A-scans per second was demonstrated. Interchangeable water immersion objectives with magnifications of 10×, 20×, and 40× provided co-registered en face cellular-resolution imaging over several size scales. Volumetric OCM data sets and en face OCM images were demonstrated on both normal and pathological human colon and kidney specimens ex vivo with an axial resolution of ~4.2 µm, and transverse resolutions of ~2.9 µm (10×), ~1.7 µm (20×), and ~1.1 µm (40×) in tissue. In addition, en face OCM images acquired with high numerical aperture over an extended field-of-view (FOV) were demonstrated using image mosaicking. Comparison between en face OCM images among different transverse and axial resolutions was demonstrated, which promises to help the design and evaluation of imaging performance of Fourier domain OCM systems at different resolution regimes.
©2013 Optical Society of America
Optical coherence tomography (OCT) enables real time, three-dimensional, high-resolution imaging of intact tissue and specimens . While OCT can achieve extremely high axial resolution , the transverse resolution is not sufficient to reveal cellular or subcellular features. Optical coherence microscopy (OCM) combines low coherence detection with conventional confocal microscopy to improve the transverse resolution of OCT images [3,4]. In confocal microscopy, a high numerical aperture (NA) optical design is essential to provide fine axial sectioning capability and reject unwanted light from out-of-focus regions. Confocal microscopy is therefore vulnerable to aberration and multiple scattering in biological specimens, which typically limits the imaging depth. In comparison, OCM significantly enhances the maximum imaging depth in scattering materials by improving rejection of scattered light using coherence gating, which also relaxes the constraint of extremely high NA optics required to obtain axial sectioning in confocal microscopy [3,4]. However, there still a trade off between the depth-of-focus (DOF) and transverse resolution in OCM images. Recent studies have used beam shaping or computational methods to demonstrate increased DOF in OCT or OCM imaging without sacrificing lateral resolution [5–13]. Although an extended DOF can be achieved using a Bessel-beam illumination and collection with an axicon lens [5–8] or binary phase filter , sensitivity losses of up to tens of decibels were reported. An extended focus OCT image was demonstrated based on digital focusing by solving the inverse scattering problem without the need for beam shaping [10–13]. However, digital focusing is computational expensive and requires phase stable acquisition along consecutive axial scans.
OCM was originally developed using time domain detection, which allows en face imaging of scattering media . Recently, an integrated OCT and OCM system was demonstrated, providing images with an axial resolution of <4 µm [14,15]. OCT enables imaging with a larger field-of-view (FOV) (3 mm × 1.5 mm) which can be used to identify regions of interest, while OCM achieves cellular-resolution imaging at the expense of a smaller FOV (400 µm × 400 µm), enabling the examination of finer tissue architectural changes in different pathologies. Integrated OCT/OCM has been utilized to image normal and diseased tissue in the human gastrointestinal tract, thyroid, breast, and kidney [15–18]. However, time domain systems are sensitive to path length variation between the coherence gate and confocal gate, which can be introduced by focal depth variation as well as perturbations to the optical fibers. This necessitates reference path adjustment to optimize signal strength during time domain OCM imaging, particularly for endoscopic applications . This requirement also makes it difficult to integrate OCM into standard scanning microscope instruments because most scanning designs introduce path length variation during scanning. Furthermore, the use of objectives with different magnifications is complicated by the need to precisely match dispersion between the reference and sample arms.
The recent development of Fourier domain detection has enabled OCT systems with 10 to 100 fold improvements in detection sensitivity and imaging speed when compared with time domain detection [19–21]. In contrast, for OCM, time domain and Fourier domain detection sensitivities are comparable given the same exposure time because only detection of light from a single en face plane at a given depth is typically performed . However, if Fourier domain detection is used for OCM applications, it demands extremely high voxel rates because en face images must be constructed from three-dimensional volumes with complete axial scans. Although Fourier domain detection has improved imaging speed for cross sectional images, the speed for en face imaging is still limited by the need to acquire an entire volume. Nonetheless, current state-of-the-art Fourier domain systems can operate at axial scan rates of several hundred kHz to more than one MHz, potentially enabling megapixel en face images to be acquired at speeds comparable to time domain OCM systems [23–26].
More importantly, Fourier domain detection enables simultaneous acquisition of all depths around the focal plane, which allows reconstruction of en face images at multiple depths and compensation of the variation between coherence gate depth position and confocal depth position. Furthermore, OCM volumetric imaging allows correction of field curvature and coherence gate curvature which is critical for high NA OCM systems with non-telecentric optical designs where the pivot of the scanning mechanism is not perfectly positioned at the back focal plane of the scan lens . This will enable OCM to be integrated into a wide range of existing scanning microscopes. Finally, Fourier domain detection enables simple numerical dispersion compensation using multiple objectives. Therefore, objectives can be easily interchanged to obtain co-registered images with different FOV.
Both spectral-domain OCM (SD-OCM) and swept-source OCM (SS-OCM) have been demonstrated in several studies [6,22,28–35]. However, given the bandwidth and center wavelength constraints of currently available swept source lasers, SD-OCM provides a crucial resolution advantage compared with SS-OCM. Using a multiplexed superluminescent diode (SLD) light source, images with axial resolution of <3 µm have been demonstrated in ophthalmic applications using spectral-domain OCT systems [36,37]. Recently, using a broadband supercontinum generation light source, an axial resolution of ~1 µm in human coronary tissue ex vivo has been achieved . However, supercontinum light sources require special attention to cancel background noise due to temporal fluctuations of the spectrum. In this paper, we report ex vivo cellular images acquired from human tissues with a compact SD-OCM system operating at an imaging speed of 210,000 A-scans per second using a new 512-pixel line scan camera (e2V Aviiva EM4). The system is demonstrated for rapid, high resolution, multiscale imaging of tissue specimens. En face images of normal and diseased colon and kidney are demonstrated using 10×, 20×, and 40× objectives. Furthermore, representative extended-FOV OCM images were demonstrated using image mosaicking.
2. Experimental setup
2.1 Spectral-domain OCM system
Figure 1 shows a schematic diagram of the SD-OCM system. A commercially available miniature SLD light source (Superlum D840) and a new high speed CCD line scan camera (e2V Aviiva EM4) were used to develop an ultrahigh speed SD-OCM system at 840 nm with an imaging speed of 210,000 A-scans per second. Although the speed was lower than that of some high speed CMOS cameras, the CCD camera had the advantage of low noise, high sensitivity and high dynamic range . The full width at half maximum (FWHM) bandwidth of the SLD light source was ~98 nm with a center wavelength at 840 nm (Fig. 2(a)). The spectrometer consisted of a collimating lens with an effective focal length of 51 mm, a 600 lines/mm transmission holographic grating, and a 100 mm focal length scan lens. The line scan camera had 512 pixels with pixel size of 14 µm × 14 µm, and the exposure time was 4.5 µs at a line rate of 210 kHz. The spectrometer was calibrated following a method reported by Makita et al.  in which reflections at two calibrated delays were used to determine the spectral scale on the CCD. The raw output signal was processed by spline interpolation using a calibration trace, followed by numerical dispersion compensation, spectral shaping using a Hamming window (Fig. 2(b)), and fast Fourier transformation (FFT). To increase the number of pixels for each A-line, a zero-padded 2048-pixel FFT was applied.
The output power of the light source after the isolator was measured to be ~7 mW and the incident power on the sample was ~2.5 mW, yielding an average sensitivity of 95.6 dB. A 50/50 coupler was used to interfere the sample and reference arm light. The total axial imaging range was measured to be ~470 µm in air (Figs. 2(d) and 2(e)) with an axial resolution of ~4.5/~5.8 µm (without/with spectral shaping) in air (Fig. 2(c)), corresponding to ~3 /~4.2 µm (without/with spectral shaping) in tissue. The imaging depth was sufficient for microscopy imaging applications because the signal range of interest is limited by the depth of focus from the confocal gate. The sample arm of the SD-OCM system was interfaced to a scanning confocal microscope. Objectives with 10×, 20×, and 40× magnifications were mounted on a turret, thereby enabling imaging with different FOV. Detailed information on the microscope can be found in section 2.2.
A polarization controller and an iris were used to set reference arm power and optimize system sensitivity. Although information on the glass in the objectives was not available, a prism pair using SFL6 glass was used to partially compensate dispersion for the 40× objective. With the SFL6 glass prism pair in the reference arm, residual dispersion imbalances from the 40× and the 10× or the 20× objectives were compensated numerically by post-processing. In addition to the dispersion, the optical path length in the sample arm varied when switching between objectives. Therefore, a linear translation stage was used to allow fast adjustment of optical path length differences up to ~1 cm.
2.2 Scanning fiber-optic confocal microscope
The fiber-optic coupled scanning confocal microscope consisted of a pair of closely spaced galvanometer-controlled scanning mirrors (Cambridge Technology 6210H, 5 mm mirrors) to provide two-dimensional scanning (Fig. 1). The fiber output was collimated onto the scanner using a near-IR achromatic lens (f = 18 mm) and relayed to the objective by a pair of identical near-IR achromatic doublet lenses (f1 = f2 = 75 mm). Water immersion objectives with three different magnifications (10× Zeiss W-Achroplan 440039, 20× Zeiss W-N-Achroplan 420957, and 40× Zeiss W-Achroplan 440095) were turret mounted to allow rapid interchange of the magnifications and FOVs while imaging. The back aperture of the objectives was under-filled and thus the effective NA was lower than that of the objective specifications.
Using antireflection coated components for the near-IR regime, a high single-pass throughput of ~80% was achieved when using the 10× and 20× objectives, and ~78% for the 40× objective. Figure 3(a) shows the coherence-gated image of a U.S Air Force resolution target (element 3 bars in group 8 can be clearly observed, inset in Fig. 3(a)) acquired using the 20× objective. Figure 3(b) shows the measurement of the coherence gate function (spectrally shaped point spread function) and confocal gate function of the 10×, 20×, and 40× objectives, respectively. The confocal gate function was measured by recoding the dc photodetector signal of the back-coupled light power from a silver mirror while translating the mirror in the sample arm around the focus position with a precision motorized stage. Distilled water was used as the immersion medium while measuring the confocal gate function. The coherence gate function (PSF) was rescaled to be in water to use the same scale as confocal gate function.
To further characterize the transverse resolution of the different objectives, a knife-edge measurement was performed. The Gaussian spot e−2 radius was calculated from the knife edge measurement to be 2.43 µm, 1.47 µm, and 0.94 µm for the 10×, 20×, and 40× objectives, respectively. A summary of the 10×, 20×, and 40× objectives is provided in Table 1. The transverse resolution decreased proportionally to 1/NA, while confocal axial resolution (DOF) decreased proportionally to 1/NA2. In addition, because the back aperture of the objectives was underfilled, the effective NAs for the 10×, 20×, and 40× objectives are also listed.
2.3 Correction of image distortion
In this paper, the goal was to design a compact SD-OCM system for high-speed and high-resolution imaging of either surgical resection or biopsy specimens. This type of instrument could also be used for a wide range of biological or materials microscopy applications. A non- telecentric optical scan design with a pair of closely spaced galvo mirrors was used to achieve a compact scan geometry. This design is similar to that used in many scanning microscopes and has a path length variation as the beam is scanned, which results in artifacts in cross-sectional and en face images. Therefore it is difficult to adapt time domain OCM to conventional microscopes. SD-OCM overcomes this limitation and can be used with a wide range of microscope designs. By taking advantage of the volumetric imaging capability of spectral-domain detection, curvature artifacts observed in the cross-sectional or en face images can be corrected in post processing.
A ~1 mm thick microscope slide was placed over the specimen to create a flat imaging surface as well as to reduce the parasitic interference signal between the specimen and the two surfaces of the slide. A thick slide was used because it moves the glass to water interface away from the objective focus. However it is important to note that a standard, thin microscope cover glass could be used if it were antireflection coated to reduce reflections from the glass to water interface. The 1 mm slide also introduced refraction artifacts, which resulted in additional curvature artifacts in the cross sectional images. Figure 4(a) shows a representative cross sectional image of a normal human kidney specimen using the 20× objective. Due to the field curvature, the interface between the specimen and the bottom surface of the slide is observed as a curved white line. The focus position within the specimen across the imaging field is marked as a red-dashed line (Fig. 4(a)). As a result of the refraction artifacts, the curved white line is not parallel with the focus position (red-dashed line). To correct the image distortion as a result of field curvature and refraction artifacts, a two step correction method was applied. First, the interface between the specimen and the slide was detected for every B-scan. Then, a two-dimensional polynomial fit was applied over the surface map generated in the previous step to remove any detection errors. Finally, every A-scan was shifted based on the fitted surface map to generate field-curvature-corrected cross sectional images (Fig. 4(b)). The variable focus position within the specimen is still evident in the field-curvature-corrected cross sectional image due to refraction artifacts introduced by the 1 mm thick microscope slide. To characterize the resolution, we measured the spot using knife-edge analysis with and without the 1-mm thick slide between the objective and resolution target. The transverse resolution was degraded from 0.78 µm to 0.94 µm for the 40× objective and from 1.38 µm to 1.47 µm for the 20× objective. The 10× objective was not measured since resolution degradation a lower NA would be even smaller. These measurements showed the loss of resolution from using the 1 mm slide was not significant.
Figures 4(c) and 4(d) shows en face OCM images at two different depth levels from the field-curvature-corrected volumetric data set, which shows changes in signal intensity across the field due to remaining refraction artifacts. To further correct the refraction artifacts, the focus position across the entire imaging field was detected by evaluating image sharpness to generate the final distortion-corrected en face OCM images. First, the curvature-corrected volumetric data set was cut into multiple small sub-volumetric data sets across the en face imaging plane. Then, the sharpness was evaluated by calculating the intensity gradient of the en face image along the depth for each sub-volumetric data set. Next, the depth position showing maximum intensity gradient was set as the focus position for each sub-volumetric data set. Finally, the distortion-corrected volumetric data set was generated by shifting each sub-volumetric data set with respect to the focus position identified individually. Figure 4(e) shows distortion-corrected en face OCM image with no significant intensity variations due to field curvature or remaining refraction artifacts observed.
2.4 Imaging protocol
2.4.1 Specimen preparation and imaging procedures
The imaging protocol was approved by the institutional review boards at Beth Israel Deaconess Medical Center (BIDMC) and Massachusetts Institute of Technology (MIT). Discarded fragments of surgically excised human specimens (colon and kidney) not requiredfor diagnosis were imaged. Specimens (typically ~1.0 × ~1.0 × ~0.5 cm3) were preserved in RPMI 1640 medium (Invitrogen) and imaged within 2 to 6 hours after surgical resection. The tissue specimens were mounted inside a modified tissue embedding cassette (Leica Biosystems) where a window was cut out on one side of the cassette and a microscope slide was used to create a flat imaging surface. The embedding cassette helped preserve the orientation of the specimen, simplifying the registration between en face OCM images and histological sections. The specimen was first imaged using the 10× objective to rapidly survey the surface of the specimen to identify regions of interest (ROIs). Multiple volumetric data sets consisting of 800 × 800 A-scans (1.76 × 1.76 mm2, X × Y) were acquired after adjusting the focus position beneath the tissue surface. Once the ROIs were identified, higher magnification objectives (20× or 40×) were used to image the ROIs and examine features with higher transverse resolutions. The imaging area for the 20× and 40× objectives were 900 × 900 µm2 (X × Y) and 440 × 440 µm2 (X × Y), respectively. Each volumetric data set consisted of 800 × 800 A-scans. After imaging, the specimen (including the tissue cassette) was fixed in 10% neutral buffered formalin for a minimum of 24 hours before histological processing.
2.4.2 Image processing and analysis
En face OCM images (800 × 800 pixels, X × Y) as well as cross-sectional images (800 × 1024 pixels, X × Z) were reconstructed from the distortion-corrected volumetric data set, each acquired in ~3 seconds. For each en face OCM image, spatial filtering with a 3x3 triangular kernel was applied for speckle reduction prior to square-root intensity compression and contrast adjustment. All en face and cross-sectional OCM images are displayed with an inverted gray scale where hyperscattering and hyposcattering regions are represented as black and white, respectively. The square-root compression offered a compromise between preserving the dynamic range of the image while also accentuating contrast. The cross-sectional images are displayed in standard logarithmic intensity scale.
2.4.3 Extended field-of-view imaging
Both the FOVs of the 20× and 40× objectives were limited to less than 1 mm, which was much smaller than the typical surface area of the specimen. To extend the FOV using relatively high NA objectives, image mosaicking was performed . Partially overlapping en face OCM images with high transverse resolution covering a centimeter-square FOV were acquired at the cost of increased image acquisition time. Figure 5 shows the image mosaicking scheme using en face OCM images of a normal human kidney specimen imaged with the 20× objective. The specimen was translated with a raster pattern (Fig. 5(a)) while individual volumetric data sets were acquired. Each volumetric data set was acquired with an overlap ratio of 33%, and each individual reconstructed en face OCM image was fused together using Image Composite Editor (ICE) (Microsoft Research) to generate an extended-FOV image (Fig. 5(b)) with high transverse resolution. The mosaic imaging mode provided a solution to the trade-off between the FOV and transverse resolution, which is especially useful when the size of the ROI is larger than the FOV of the higher NA objectives. Nine en face OCM images extracted from individual volumetric data sets (800 × 800 A-scans) were mosaicked to generate a large FOV image where the total acquisition time was ~30 seconds.
Normal and diseased human colon and kidney specimens were imaged using the compact ultrahigh speed SD-OCM system. Representative en face OCM images at different magnifications are displayed. Images displayed at one magnification scale can be co-registered to images at the other two magnification scales. In addition, en face OCM images with extended FOV are shown.
3.1 Human colon
Figure 6 shows co-registered en face OCM images of a normal human colon specimen at different magnifications. Normal human colon is lined by the colonic mucosa, which is rich in goblet cells. Colonic epithelium covering the mucosa invaginates into the lamina propria at regular intervals forming structures called crypts. Using the 10× objective, the en face OCM image (Fig. 6(a)) shows a regular pattern of crypts across the field. With limited transverse resolution, goblet cell morphology is not readily identified in the 10× en face OCM image (red arrows, Fig. 6(a)). Figure 6(b) shows a 20× en face OCM image over the ROI (green dashed box) selected in Fig. 6(a). With improved transverse resolution, the morphology of the crypts and goblet cells are better visualized, including the crypt lumen and epithelium. The 40× image in Fig. 6(c) further enhances the goblet cell visibility and reduces speckle-noise.
Figure 7 shows representative en face OCM images extracted from a single volumetric data set using the 40× objective. Figure 7(c) shows the en face OCM image at the focus position while Figs. 7(a) and 7(e), and Figs. 7(b) and 7(d) show en face OCM images extracted at locations 13.6 µm and 6.8 µm above and below the focus position, respectively. Figures 7(f)-7(j) shows en face OCM images over the ROI (red dashed boxes) highlighted in Figs. 7(a)-7(e), respectively. Due to short DOF of the 40× objective, the morphology of thecrypts and the contour of the goblet cells become blurred in the extracted en face images located at 13.6 µm away from the focus position (Figs. 7(a) and 7(f)). Nevertheless, as shown in Fig. 7, although a high NA objective was used, multiple en face OCM images could still be extracted from a single volumetric data set within a reasonable axial range around the focus slightly longer than the DOF of 40× objective. By taking multiple en face OCM images from a single volumetric data set, the effective imaging rate can be increased. In comparison, Fig. 8 shows representative en face OCM images extracted from a single volumetric data set using the 10× objective. Due to the use of lower NA of the 10× objective, the DOF is greatly increased, supporting en face imaging at depths up to ~42.7 µm away from the focus. Therefore, using the 10× objective, ROIs can be identified not only in the x-y en face imaging plane but also along the axial direction. Once the 3D location of the ROI is identified, a high-precision 3D motorized stage is utilized so that fine structure within the ROI can be examined using objectives with higher magnifications, e.g. the 20× and 40× objectives.
Figure 9 shows high-resolution, cellular-level images of crypt structures in a human colon specimen at different focus positions beneath the tissue surface. Multiple volumetric OCM data set were acquired by gradually translating the focus position deeper into the tissue with 10 µm increments using the 40× objective. Only the in-focus en face OCM images were selected from individual volumetric data sets and displayed as a function of the focus position. As marked by the arrows in each en face OCM image, the size of crypt lumen becomes narrower in en face images at deeper focus position. For the en face image at 95 µm, the imaging plane was close to the bottom of the crypt.
Figure 10 shows co-registered multiscale OCM images of a human colon specimen with ulcerative colitis (UC), which is a form of inflammatory bowel disease (IBD). Because of active and chronic inflammation and ulceration, UC is characterized by severe architectural distortion of the crypts and expansion of lamina propria by inflammatory infiltrate. Imaged with the 10× objective (Fig. 10(a)), the crypt density is lower than that of normal human colon (Fig. 7(a)). In addition, the crypts have irregular size and shape. Some large, dilated, and distorted crypts can be observed (arrows, Fig. 10(a)). Using the 20× and 40× objective, visibility of the morphology of the distorted crypts is improved (Figs. 10(b) and 10(c)). Absence of goblet cells is typical. Figure 10(d) shows the corresponding H&E histopathology confirming the findings in the en face OCM images.
Figure 11 shows an extended-FOV en face OCM image and the corresponding H&E histopathology image of the same specimen as Fig. 10, using the 20× objective. Nine individual en face OCM images with an overlap ratio of 33% were used to generate the mosaicked OCM image over a large imaging area (2.1 × 2.1 mm2, X × Y), slightly larger than that of the 10× objective. In comparison with the 10× en face OCM image (Fig. 10(a)), the visibility and image contrast over the region with distorted crypts are improved, and the FOV was improved from 900 × 900 µm2 for an individual 20× OCM image to 2.1 × 2.1 mm2 in the mosaicked image. Although an extended FOV can be realized, it is not efficient to survey the surface of the specimen using image mosaicking under a relatively high NA objective due to the increased imaging time. However, it is possible to achieve wide FOV and fine transverse resolution using image mosaicking to fill the gap between 10× and 20× objective images.
3.2 Human kidney
Figure 12(a) shows the en face OCM image of normal human kidney specimen. Characteristic structures of normal human kidney in the cortex including the round, spherical glomerulus (G) and surrounding convoluted tubules (T) can be observed. However, due to the limited transverse resolution, the fine structures inside the glomerulus (e.g. capillary loops and mesangial cells) cannot be clearly identified. Therefore, the ROI region (green dashed box, Fig. 12(a)) was zoomed in using the 20× objective. As shown in Fig. 12(b), the image FOV (900 × 900 µm2, X × Y) was effectively decreased in half, while the resolution, contrast and speckle noise are improved. The branching vascular network in the glomerulus is characterized by round, pale regions surrounded by a hyposcattering ring, representing Bowman’s space. Convoluted tubules are characterized as hyperscattering rings surrounded by hyposcattering stroma. To further examine the glomerulus, the ROI (blue round dotted box, Fig. 12(b)) was imaged using the 40× objective. Under higher magnification, hyperscattering structures reminiscent of podocytes inside Bowman's space (arrows, Fig. 12(c)) can be observed more clearly.
Figure 13 shows the cross sectional OCM images of the normal human kidney specimen using the 10×, 20×, and 40× objectives. As a result of the increased transverse resolution, a decrease in speckle sizes can be observed in the cross sectional images. In addition, compared with the 10× cross sectional image, the focus position inside the specimens can be clearly identified in the 20× and 40× cross sectional image since the DOFs are shorter and thus the DOF region is highlighted when using higher NA objectives.
Figure 14 shows co-registered multiscale OCM images of a human kidney specimen with oncocytoma. Histopathologically, oncocytoma is characterized by oxyphilic tumor cells arranged in solid nests or cyst/tubular structures embedded in surrounding myxoid or hyalinized stroma. Under the 10× objective (Fig. 14(a)), characteristic components of normal human kidney, including glomeruli and convoluted tubules are replaced by tumor cell clusters. Several round tumor nests are observed (arrows, Fig. 14(a)), while the tumor cells in the upper right part of the image are more packed and thus no clear island-like growth pattern exists. To improve the contrast and further examine the tumor features at the cellular level, both the 20× and 40× objectives were used. As shown in Fig. 14(b), the ROI showing tumor nests surrounded by less myxoid stroma (blue round dotted box) can be identified. Using the 40× objective, the tumor nests and surrounding myxoid stroma are more clearly differentiated (Fig. 14(c)).
Figure 15 shows an extended-FOV en face OCM image and the corresponding histopathology image of the same specimens as Fig. 14 using the 20× objective. Following the same image mosaicking protocol as Fig. 11, an imaging field of 2.1 × 2.1 mm2 was acquired. The imaging area was close to that of Fig. 14, slightly shifted along the Y (vertical) direction. In comparison with Figs. 5 and 12, solid round tumor nests, one of the characteristic features of oncocytoma, can be observed with improved contrast (red arrows, Fig. 15(a)).
Figure 16 shows a comparison of the en face OCM images as a function of both transverse and axial resolution. Objectives with three different magnifications (10×, 20×, and 40×) were used to acquire en face OCM images with different transverse resolutions, while different window sizes were selected for spectral shaping during processing to vary the axial resolution from 4.2 µm to 8.4 µm and 16.8 µm in tissue. As a result of the lower transverse resolution, the 10× objective en face image features are initially less sensitive to the loss of axial resolution (spectrally shaped axial resolution in tissue (a) 4.2 µm, and (b) 8.4 µm). However, when the axial resolution is degraded further ((c) 16.8 µm) the image contrast decreases. The image contrast will continue to decrease if the axial resolution is further degraded. In comparison, with higher NA and finer transverse resolution (20×, and 40×), the contrast of the en face image features decreases with the initial degradation of axial resolution where the contours of the convoluted tubules (red arrow, Fig. 16(f)) and glomerulus (red arrow, Fig. 16(i)) become blurred. However, in the high NA focusing case, the confocal gate rejects unwanted scattered light and as the coherence gate function is made wider, the image contrast will not change appreciably. In comparison to the 20× objective, the transverse resolution of 40× objective is very high and thus the changes in image contrast is observed most clearly (red arrows, Figs. 16(h)-16(i)) even though the DOF confocal gate function is short. In both low and high transverse resolution cases, the changes in image contrast also depend on the structure being imaged. If the structure has a rapid variation in features with depth, the contrast and resolution are lost faster when more unwanted scattered light is detected.
4. Discussion and conclusion
We have demonstrated a co-registered, multiscale OCM system for imaging freshly excised tissue specimens. Representative ex vivo OCM images of both normal and diseased human kidney and colon tissue at three different resolution scales showed features associated with pathologies without the use of exogenous contrast agents. The system can be used in a manner similar to the way a pathologist uses a conventional light microscope to analyze histopathology slides, seamlessly changing between magnifications to integrate feature analysis of tissue architecture and cellular morphology.
Using the lower magnification objective (10×), the system provided architecture-level imaging of gross tissue pathologies over a relatively large FOV (1.76 × 1.76 mm2), and thus enabling rapid survey of specimens to identify ROIs. Once the ROIs are identified, higher magnification objectives (20× and 40×) were used to further examine the ROIs with cellular resolution but with reduced FOV (20×: 900 × 900 µm2; 40×: 440 × 440 µm2). Under higher magnification objectives, volumetric imaging at different focus depths and image mosaicking were performed. This multiscale imaging approach enables efficient survey of entire specimens and facilitates high throughput imaging of clinical samples.
Compared with previous time domain based OCM systems [14,15], the system complexity was reduced because path length delay variations with scanning could be compensated in post processing. Therefore it was not necessary to design a path length invariant beam scanning module or to actively track the coherence gate delay to match the confocal gate delay. The ability to perform numerical dispersion compensation allows a simple reference arm design, because adjustable dispersion compensating glass is not required. This allows rapid switching of objectives with different magnifications in the sample arm. These features of Fourier domain detection enable easier integration of OCT/OCM with other microscopy instruments, such as two-photon microscopy [30,32,33,40] and laser-scanning fluorescence microscopy [41–43]. Furthermore, SD-OCM can provide either cross-sectional or en face phase information from the specimen [28,29].
When compared with other phase microscopy techniques, such as phase contrast and differential interferential interference contrast microscopy, SD-OCM can provide quantitative analysis of phase information of materials or biological specimens. Using polarization-sensitive detection, SD-OCM can provide birefringence analysis including retardation, axis of retardation and degree of polarization non-uniformity.
By varying the axial resolution using different spectral shaping windows, this system was used to investigate image quality as a function of both transverse and axial resolution determined by the NA of focusing optics and the light source bandwidth. As shown in Fig. 16, the combination of both higher NA focusing optics and shorter coherence gate function (the PSF width) allows OCM imaging with better image quality. Additionally, we can evaluate design parameters to optimize both the transverse and axial resolutions for different imaging applications.
For example, as shown in Figs. 6 and 10 and also in previous studies [15,44,45], crypt morphology patterns were correlated with colon pathologies. OCM promises to facilitate the detection of dysplasia or differentiation between hyperplasic and adenomatous polyps in surveillance colonoscopy. However, the imaging performance of an endoscopic OCM system design must be carefully evaluated prior to in vivo imaging. This study demonstrated an OCM system capable of quickly assessing the imaging quality in specimens ex vivo with parameters similar to those which would be used in vivo. This approach can aid the design of OCM systems in general by testing system parameters.
The en face imaging rate of the SD-OCM system is limited by the acquisition speed of the line scan camera at 210,000 A-scans per second. Although time domain OCM can have a very high en face imaging rate (~700,000 pixels per second in prior implementation [14,15]), SD-OCM acquires volumetric data sets where multiple en face OCM images can be extracted. Compared with typical SS-OCM systems , the SD-OCM system supports volumetric imaging with a higher axial resolution of ~4.2 µm in tissue. The higher axial resolution provides improved image contrast and decreased speckle size.
In summary, we developed a ultrahigh speed, high resolution, and compact SD-OCM system supporting multiscale imaging of tissue pathologies ex vivo. Performing OCM imaging at different resolution scales is analogous to the way pathologists use multiple magnifications during microscopic examination of a pathologic specimen. As a result of the volumetric imaging capability, the image path delay distortion can be corrected in post processing. Given the high imaging speed with cellular resolution and compact system design, this system is well suited for imaging tissue specimens intraoperatively in a pathology laboratory as well as for a range of other biological and materials applications.
We gratefully thank Drs. Yuankai K. Tao, Michael G. Giacomelli and Ireneusz Grulkowski for scientific discussions and study design. We also like to thank Ilias Levis of Ikona Inc. for providing e2V test cameras. The research was sponsored in part by the National Institutes of Health R01-CA75289-16, R44-CA101067-06, R01-EY011289-27, R01-HL095717-04 and R01-NS057476-05; the Air Force Office of Scientific Research FA9550-12-1-0499 and Medical Free Electron Laser Program FA9550-10-1-0551.
References and links
1. D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G. Stinson, W. Chang, M. R. Hee, T. Flotte, K. Gregory, C. A. Puliafito, and J. G. Fujimoto, “Optical coherence tomography,” Science 254(5035), 1178–1181 (1991). [CrossRef] [PubMed]
2. W. Drexler, U. Morgner, R. K. Ghanta, F. X. Kärtner, J. S. Schuman, and J. G. Fujimoto, “Ultrahigh-resolution ophthalmic optical coherence tomography,” Nat. Med. 7(4), 502–507 (2001). [CrossRef] [PubMed]
4. M. Kempe, W. Rudolph, and E. Welsch, “Comparative study of confocal and heterodyne microscopy for imaging through scattering media,” J. Opt. Soc. Am. A 13(1), 46–52 (1996). [CrossRef]
5. Z. Ding, H. Ren, Y. Zhao, J. S. Nelson, and Z. Chen, “High-resolution optical coherence tomography over a large depth range with an axicon lens,” Opt. Lett. 27(4), 243–245 (2002). [CrossRef] [PubMed]
6. R. A. Leitgeb, M. Villiger, A. H. Bachmann, L. Steinmann, and T. Lasser, “Extended focus depth for Fourier domain optical coherence microscopy,” Opt. Lett. 31(16), 2450–2452 (2006). [CrossRef] [PubMed]
7. K. S. Lee and J. P. Rolland, “Bessel beam spectral-domain high-resolution optical coherence tomography with micro-optic axicon providing extended focusing range,” Opt. Lett. 33(15), 1696–1698 (2008). [CrossRef] [PubMed]
8. C. Blatter, B. Grajciar, C. M. Eigenwillig, W. Wieser, B. R. Biedermann, R. Huber, and R. A. Leitgeb, “Extended focus high-speed swept source OCT with self-reconstructive illumination,” Opt. Express 19(13), 12141–12155 (2011). [CrossRef] [PubMed]
9. L. Liu, C. Liu, W. C. Howe, C. J. R. Sheppard, and N. Chen, “Binary-phase spatial filter for real-time swept-source optical coherence microscopy,” Opt. Lett. 32(16), 2375–2377 (2007). [CrossRef] [PubMed]
10. T. S. Ralston, D. L. Marks, F. Kamalabadi, and S. A. Boppart, “Deconvolution methods for mitigation of transverse blurring in optical coherence tomography,” IEEE Trans. Image Process. 14(9), 1254–1264 (2005). [CrossRef] [PubMed]
11. L. F. Yu, B. Rao, J. Zhang, J. P. Su, Q. Wang, S. G. Guo, and Z. P. Chen, “Improved lateral resolution in optical coherence tomography by digital focusing using two-dimensional numerical diffraction method,” Opt. Express 15(12), 7634–7641 (2007). [CrossRef] [PubMed]
12. Y. Yasuno, J. I. Sugisaka, Y. Sando, Y. Nakamura, S. Makita, M. Itoh, and T. Yatagai, “Non-iterative numerical method for laterally superresolving Fourier domain optical coherence tomography,” Opt. Express 14(3), 1006–1020 (2006). [CrossRef] [PubMed]
13. T. S. Ralston, D. L. Marks, P. S. Carney, and S. A. Boppart, “Interferometric synthetic aperture microscopy,” Nat. Phys. 3(2), 129–134 (2007). [CrossRef]
14. A. D. Aguirre, J. Sawinski, S.-W. Huang, C. Zhou, W. Denk, and J. G. Fujimoto, “High speed optical coherence microscopy with autofocus adjustment and a miniaturized endoscopic imaging probe,” Opt. Express 18(5), 4222–4239 (2010). [CrossRef] [PubMed]
15. A. D. Aguirre, Y. Chen, B. Bryan, H. Mashimo, Q. Huang, J. L. Connolly, and J. G. Fujimoto, “Cellular resolution ex vivo imaging of gastrointestinal tissues with optical coherence microscopy,” J. Biomed. Opt. 15(1), 016025 (2010). [CrossRef] [PubMed]
16. C. Zhou, Y. H. Wang, A. D. Aguirre, T. H. Tsai, D. W. Cohen, J. L. Connolly, and J. G. Fujimoto, “Ex vivo imaging of human thyroid pathology using integrated optical coherence tomography and optical coherence microscopy,” J. Biomed. Opt. 15(1), 016001 (2010). [CrossRef] [PubMed]
17. C. Zhou, D. W. Cohen, Y. Wang, H.-C. Lee, A. E. Mondelblatt, T.-H. Tsai, A. D. Aguirre, J. G. Fujimoto, and J. L. Connolly, “Integrated optical coherence tomography and microscopy for ex vivo multiscale evaluation of human breast tissues,” Cancer Res. 70(24), 10071–10079 (2010). [CrossRef] [PubMed]
18. H.-C. Lee, C. Zhou, D. W. Cohen, A. E. Mondelblatt, Y. Wang, A. D. Aguirre, D. Shen, Y. Sheikine, J. G. Fujimoto, and J. L. Connolly, “Integrated optical coherence tomography and optical coherence microscopy imaging of ex vivo human renal tissues,” J. Urol. 187(2), 691–699 (2012). [CrossRef] [PubMed]
19. J. F. de Boer, B. Cense, B. H. Park, M. C. Pierce, G. J. Tearney, and B. E. Bouma, “Improved signal-to-noise ratio in spectral-domain compared with time-domain optical coherence tomography,” Opt. Lett. 28(21), 2067–2069 (2003). [CrossRef] [PubMed]
21. M. A. Choma, M. V. Sarunic, C. H. Yang, and J. A. Izatt, “Sensitivity advantage of swept source and Fourier domain optical coherence tomography,” Opt. Express 11(18), 2183–2189 (2003). [CrossRef] [PubMed]
22. S.-W. Huang, A. D. Aguirre, R. A. Huber, D. C. Adler, and J. G. Fujimoto, “Swept source optical coherence microscopy using a Fourier domain mode-locked laser,” Opt. Express 15(10), 6210–6217 (2007). [CrossRef] [PubMed]
23. B. Potsaid, I. Gorczynska, V. J. Srinivasan, Y. L. Chen, J. Jiang, A. Cable, and J. G. Fujimoto, “Ultrahigh speed spectral / Fourier domain OCT ophthalmic imaging at 70,000 to 312,500 axial scans per second,” Opt. Express 16(19), 15149–15169 (2008). [CrossRef] [PubMed]
24. W. Wieser, B. R. Biedermann, T. Klein, C. M. Eigenwillig, and R. Huber, “Multi-megahertz OCT: High quality 3D imaging at 20 million A-scans and 4.5 GVoxels per second,” Opt. Express 18(14), 14685–14704 (2010). [CrossRef] [PubMed]
25. V. Jayaraman, G. D. Cole, M. Robertson, A. Uddin, and A. Cable, “High-sweep-rate 1310 nm MEMS-VCSEL with 150 nm continuous tuning range,” Electron. Lett. 48(14), 867–868 (2012). [CrossRef]
26. L. An, P. Li, T. T. Shen, and R. K. Wang, “High speed spectral domain optical coherence tomography for retinal imaging at 500,000 A-lines per second,” Biomed. Opt. Express 2(10), 2770–2783 (2011). [CrossRef] [PubMed]
30. B. W. Graf, Z. Jiang, H. H. Tu, and S. A. Boppart, “Dual-spectrum laser source based on fiber continuum generation for integrated optical coherence and multiphoton microscopy,” J. Biomed. Opt. 14(3), 034019 (2009). [CrossRef] [PubMed]
32. B. W. Graf and S. A. Boppart, “Multimodal in vivo skin imaging with integrated optical coherence and multiphoton microscopy,” IEEE J. Sel. Top. Quantum Electron. 18(4), 1280–1286 (2012). [CrossRef]
33. L. B. Liu, J. A. Gardecki, S. K. Nadkarni, J. D. Toussaint, Y. Yagi, B. E. Bouma, and G. J. Tearney, “Imaging the subcellular structure of human coronary atherosclerosis using micro-optical coherence tomography,” Nat. Med. 17(8), 1010–1014 (2011). [CrossRef] [PubMed]
34. K.-S. Lee, H. Zhao, S. F. Ibrahim, N. Meemon, L. Khoudeir, and J. P. Rolland, “Three-dimensional imaging of normal skin and nonmelanoma skin cancer with cellular resolution using Gabor domain optical coherence microscopy,” J. Biomed. Opt. 17(12), 126006 (2012). [CrossRef] [PubMed]
35. V. J. Srinivasan, H. Radhakrishnan, J. Y. Jiang, S. Barry, and A. E. Cable, “Optical coherence microscopy for deep tissue imaging of the cerebral cortex with intrinsic contrast,” Opt. Express 20(3), 2220–2239 (2012). [CrossRef] [PubMed]
36. B. Cense, N. Nassif, T. C. Chen, M. C. Pierce, S. H. Yun, B. H. Park, B. Bouma, G. Tearney, and J. F. de Boer, “Ultrahigh-resolution high-speed retinal imaging using spectral-domain optical coherence tomography,” Opt. Express 12(11), 2435–2447 (2004). [CrossRef] [PubMed]
37. T. H. Ko, D. C. Adler, J. G. Fujimoto, D. Mamedov, V. Prokhorov, V. Shidlovski, and S. Yakubovich, “Ultrahigh resolution optical coherence tomography imaging with a broadband superluminescent diode light source,” Opt. Express 12(10), 2112–2119 (2004). [CrossRef] [PubMed]
38. H. Helmers and M. Schellenberg, “CMOS vs. CCD sensors in speckle interferometry,” Opt. Laser Technol. 35(8), 587–595 (2003). [CrossRef]
39. S. Makita, T. Fabritius, and Y. Yasuno, “Full-range, high-speed, high-resolution 1 microm spectral-domain optical coherence tomography using BM-scan for volumetric imaging of the human posterior eye,” Opt. Express 16(12), 8406–8420 (2008). [CrossRef] [PubMed]
40. C. Vinegoni, T. Ralston, W. Tan, W. Luo, D. L. Marks, and S. A. Boppart, “Integrated structural and functional optical imaging combining spectral-domain optical coherence and multiphoton microscopy,” Appl. Phys. Lett. 88(5), 053901 (2006). [CrossRef]
41. A. R. Tumlinson, J. K. Barton, B. Povazay, H. Sattman, A. Unterhuber, R. A. Leitgeb, and W. Drexler, “Endoscope-tip interferometer for ultrahigh resolution frequency domain optical coherence tomography in mouse colon,” Opt. Express 14(5), 1878–1887 (2006). [CrossRef] [PubMed]
42. S. A. Yuan, C. A. Roney, J. Wierwille, C. W. Chen, B. Y. Xu, G. Griffiths, J. Jiang, H. Z. Ma, A. Cable, R. M. Summers, and Y. Chen, “Co-registered optical coherence tomography and fluorescence molecular imaging for simultaneous morphological and molecular imaging,” Phys. Med. Biol. 55(1), 191–206 (2010). [CrossRef] [PubMed]
43. L. P. Hariri, E. R. Liebmann, S. L. Marion, P. B. Hoyer, J. R. Davis, M. A. Brewer, and J. K. Barton, “Simultaneous optical coherence tomography and laser induced fluorescence imaging in rat model of ovarian carcinogenesis,” Cancer Biol. Ther. 10(5), 438–447 (2010). [CrossRef] [PubMed]
44. C. Pitris, C. Jesser, S. A. Boppart, D. Stamper, M. E. Brezinski, and J. G. Fujimoto, “Feasibility of optical coherence tomography for high-resolution imaging of human gastrointestinal tract malignancies,” J. Gastroenterol. 35(2), 87–92 (2000). [CrossRef] [PubMed]
45. X. Qi, Y. Pan, Z. Hu, W. Kang, J. E. Willis, K. Olowe, M. V. Sivak Jr, and A. M. Rollins, “Automated quantification of colonic crypt morphology using integrated microscopy and optical coherence tomography,” J. Biomed. Opt. 13(5), 054055 (2008). [CrossRef] [PubMed]