Optical coherence tomography (OCT) is a rapidly emerging imaging modality that can non-invasively provide cross-sectional, high-resolution images of tissue morphology in situ and in real-time. We previously demonstrated that OCT is capable of visualizing characteristic kidney anatomic structures, including blood vessels, uriniferous tubules, glomeruli, and renal capsules on a Munich–Wistar rat model. Because the viability of a donor kidney is closely correlated with its tubular morphology, and a large amount of image datasets are expected when using OCT to scan the entire kidney to provide a global assessment of its viability, it is necessary to develop automatic image analysis methods to quantify the spatially-resolved morphometric parameters such as tubular diameter to provide potential diagnostic information. In this study, we imaged the human kidney in vitro and quantified the diameters of hollow structures such as blood vessels and uriniferous tubules automatically. The microstructures were first segmented from cross-sectional OCT images. Then the spatially-isolated region-of-interest (ROI) was automatically selected to quantify its dimension. This method enables the automatic selection and quantification of spatially-resolved morphometric parameters. The quantification accuracy was validated, and measured features are in agreement with known kidney morphology. This work can enable studies to determine the clinical utility of OCT for kidney imaging, as well as studies to evaluate kidney morphology as a biomarker for assessing kidney’s viability prior to transplantation.
© 2009 Optical Society of America
Optical coherence tomography (OCT) is a high-resolution, cross-sectional, three-dimensional imaging modality that measures the echo delay of light to generate images . Although the light scattering properties of biological tissues typically limit light penetration to less than 2 mm, this imaging depth has proven sufficient to provide valuable information about tissue pathology in a number of biomedical fields including ophthalmology [2–4], cardiology [5, 6], and gastroenterology [7–10]. OCT can be interfaced with various imaging devices such as catheters, endoscopes, laparoscopes, and needles, with typical image resolutions of 1–15 µm . Therefore, OCT is a promising imaging modality to assess tissue pathologies in situ and in real time. In addition, image processing has come to play an important role in understanding the information content of biological tissues [12-15].
The versatility of OCT imaging procedures, its resolution capabilities and increased depth analysis, as compared with conventional microscopy, make OCT an ideal method for imaging the human kidney in situ. OCT imaging can provide immediate information regarding the histopathological status of the renal vasculature, tubules, and glomeruli. One potential OCT application is the evaluation of the viability of donor kidneys . Previous studies using tandem-scanning confocal microscopy  indicated that proximal tubular structure and post-transplantation renal function are closely correlated. In a recent study , living rat kidneys were observed in vivo before, during, and after an ischemic insult using OCT, which enabled the visualization and comparison of the rat kidney morphology. OCT therefore represents an exciting new approach to visualize, in real-time, the pathological changes in the living kidney in a non-invasive or minimally-invasive fashion.
In the previous study , the total volume of kidney tubules were segmented and quantified. Statistically significant changes were observed during the ischemia. However, the segmentation algorithm used in this prior study quantified the average tubular volume change only, and the spatially-resolved local morphological changes were not able to be separated and quantified. Furthermore, kidney structures are heterogeneous; therefore the difference in tubular volume is confounded by the tubular density. The previous study  suggested that the tubular lumen diameter is a more robust biomarker for kidney viability, and a decrease in lumen diameter during ischemia was observed visually by both confocal microscopy  and OCT . Therefore, it is important to develop methods to quantify the spatially-resolved tubular diameters.
The most straight-forward method to quantify the tubular diameters from OCT (or other imaging modalities) images is manual measurement using calipers or partially-automated image analysis softwares (such as ImageJ). Although accurate and reproducible measurements can be obtained in this way, an obvious drawback is the extent of user interaction required for the analysis. For instance, it requires manual selection of the region-of-interest (ROI) and the tubular wall edges on the images by the operator as the first step. This procedure is very laborious and time-consuming, which precludes the possibility of analyzing large amounts of data. This is especially challenging for OCT imaging of the kidney, since individual OCT images have a field-of-view (FOV) of several millimeters while a typical human kidney has a surface area larger than 10 cm by 10 cm. To provide an accurate assessment of the entire kidney, comprehensive OCT imaging is necessary, which would involve a large number of images from various locations of the kidney. Thus, an automatic image analysis method is critical.
Our previous work has demonstrated that hollow kidney microstructurs such as uriniferous tubules and Bowman’s space can be automatically segmented based on their different backscattering intensities . Automated segmentation of colonic crypt morphology using OCT  has also been demonstrated. In addition, automatic evaluation of diameters of a single cylindrical structure such as brachial artery has been demonstrated using B-mode ultrasonic imaging [19, 20]. However, in our application, multiple isolated tubules with various diameters and curvatures are presented in a single OCT image. To obtain the spatially-resolved morphometric information, it is necessary to separate those isolated regions for further quantification. The purpose of the present study was to develop an image processing method for automatic selection of individual ROI and quantification of the size of the hollow structures in the kidney, including renal tubules, glomeruli, and vessels. Since there are significant differences in the size and structure of human kidneys and those of rodents, we undertook these studies using human kidneys for enhanced clinical relevance. This study is a necessary step before assessing the utility of OCT in clinical evaluation of kidney viability.
2. Materials and methods
2.1 Human kidney and histology
This study protocol was approved by the Institutional Review Boards (IRB) at both the University of Maryland and Georgetown University. Four donor kidneys were obtained through the Washington Regional Transplant Consortium (WRTC). Upon arrival, the kidneys were fixed by vascular perfusion with 10% neutral formalin (through the renal artery) to preserve their renal morphology. After the OCT image acquisition, the location and direction of each scanned section were marked with ink, for subsequent standard histology processing. For conventional light microscopy, 4 µm thick sections were cut, stained with hematoxylineosin (H&E), and photographed with a Nikon Eclipse 80i (Nikon, Melville, NY) attached to a digital camera Nikon DS-Fi1 (Nikon). The micrographs were obtained for comparison with OCT images.
2.2 Optical coherence tomography (OCT) imaging
This study used a high-speed high-resolution OCT system (Thorlabs Inc., NJ, USA) using swept source/Fourier domain detection that enabled three-dimensional (3D) OCT imaging in situ. The light source was a wavelength-swept laser light source generating a 100 nm full width at half maximum (FWHM) bandwidth at 1310 nm, yielding an axial resolution of 10 µm in the tissue. The laser operated at a swept rate of 16 kHz with an average output power of 12 mW. The imaging frame rate was 30 frames per second. The transverse resolution of the system was 15 µm with 4 mW of power illuminating the sample.
Figure 1 shows the overall schematic of the OCT system used in this experiment. The inset in the lower left corner shows the imaging microscope. The output of the swept laser was split into two portions: three percent was used to generate a clock signal for triggering the sample of the OCT signal on a uniformly-spaced optical frequency grid ; the remaining ninety-seven percent of the output was equally distributed to the OCT sample and reference arms. Imaging of the human kidney sample was performed by a pair of mirrors mounted to XY scanning galvanometers (Cambridge Technology, MA, USA) and a microscope objective. The OCT imaging system’s sensitivity was 97 dB.
2.3 OCT image processing and analysis
3D OCT images of the kidney measuring 3 mm by 3 mm by 2.25 mm (512×512×512 pixels) were obtained from various locations on the human kidney samples, without contact. 3D OCT images with representative microstructures were selected and compared with corresponding conventional histology. To quantitatively evaluate the OCT images and obtain diagnostic information, image processing was performed on each individual cross-sectional (XZ or YZ plane) OCT image. Figure 2 displays a general flow chart of the automatic imaging processing procedure.
2.3.1 Overview of image analysis methods
The automatic image processing method included image segmentation, region-of-interest (ROI) selection, and image feature quantification. First, the raw OCT image data were obtained (XZ and YZ). The contour of kidney surface was identified by edge detection on each A-scan. Then the structures in the kidney (such as uriniferous tubules and blood vessels) were segmented from the kidney parenchyma based on their different backscattering intensities  (Step 1 in Fig. 2).
To accurately distinguish local changes, an image processing algorithm was used to automatically identify and separate the isolated sections (i.e. uriniferous tubules) from the segmented images to quantify the diameter of each ROI (such as individual tubules or blood vessels). The algorithm systematically filled the region to the section boundary and labeled each region with a unique index. This algorithm allowed different regions to be individually selected for further morphometrical analysis (for instance, quantifying the diameter) or to count the total number of isolated sections (Step 2 in Fig. 2). This step was essential to ensure that the diameters measured are from the selected ROI, therefore, can be color-coded and displayed in a spatially-resolved way.
In this study, we focus on the quantification of tubular (or vessel) diameter. To quantify the diameter of each isolated ROI, the corresponding boundary and skeleton were generated. As a result, the diameters of each luminal position in this ROI were calculated based the average of the shortest distances from the boundary to the skeleton. To minimize the errors due to sampling, the same analysis approaches were applied to both the XZ and YZ image cross-sections, and the final dimension was calculated by averaging the values obtained from the two cross-sections (XZ and YZ). In this way, the spatially-resolved dimensional information was obtained (Step 3 in Fig. 2) and presented in 3D.
2.3.2 Segmentation of hollow structures of kidney
A 5×5 median filter was first applied to the OCT images to reduce the background speckle noise. Segmentation is then applied to the OCT images, which subdivided an image into its constituent parts to distinguish the objects of interest and the background. In this study, we used an intensity threshold to segment the OCT images. A resulting binary image g(x,y) is defined as :
where f(x,y) was the gray level of a point (x,y) in the original OCT images. Thus the pixel (x,y) corresponding to hollow structure (such as tubules) was labeled 1 in the segmented image, whereas pixels labeled 0 corresponded to the background (kidney parenchyma). In such a way all pixels with a gray level lower than empirical value threshold (T) were extracted from the background for each image.
2.3.3 Automatic selection of isolated ROIs
The intensity values of the segmented images g(x,y) were scanned pixel by pixel. The background intensity was 0 and each isolated ROI was 1, every time a 1 was detected, the program was triggered to fill the region. The filling process flooded the region in four directions (up, down, left and right) until reaching the boundary, i.e., encountering 0. The filling process was performed in MATLAB based on the function “encodem”. Figure 3 shows the process of segmentation followed by the automatic selection of individual isolated ROIs. The filling process filled different regions with different values (as indicated by different colors in Fig. 3) so that it could count the regions and extract each region from the image for further processes.
2.3.4 Quantification of image features
After extraction of an individual ROI, further morphometric analysis could be performed on the region. In the present study, we focused on the estimation of the diameter of the tubular lumen. The diameter was quantified by measuring the radius, which was the minimal distance from a specific boundary pixel to the skeleton (see Fig. 4). The boundary was defined as a pixel set where the spatial neighbor of every member contains both intensity 1 and 0 pixels. The skeleton is another pixel set which represents spatially a minimally connected stroke that a region thins to . The boundary and skeleton were obtained by the MATLAB function “bwmorph”. By using these two pixel sets, the radius for every pixel (b) on the boundary (B) was defined to be:
A radius was determined for every point b by finding the minimal distance between b and the skeleton set (S). This process was applied to both cylindrical and non-cylindrical (c.f., with branches) structures as illustrated in Fig. 4. For visual purposes, pixels’ intensities were rendered with a number of 2 times the associated radius (the local diameter of the feature).
3.1 Calibration of the dimension calculation algorithm
To quantitatively assess the accuracy of the dimension calculation algorithm, we applied this algorithm to the dimensional calculation of a capillary tube phantom. By comparing the computer calculated results with manual measurements, the performance of the algorithm was validated. Figure 5(a) shows one representative cross-sectional OCT image (YZ) of a capillary tube phantom, with the associated segmented image shown in Fig. 5(b). Figure 5(c) shows the histogram of the automatic estimation of the tube radius from a total of 61 different YZ cross-sectional OCT images along X axis. The computer algorithm estimated the diameter of the capillary tube to be 126.6±8.6 µm. A human observer measured the diameter directly from the same set of OCT images, and result in 128.3±7.4 µm (Fig. 5(d)). The computer analysis result shows a slightly larger variance since the diameter is averaged from all boundary pixel measurements, while the human observer only select few edge pixels to quantify the diameter. Fig. 5(e) shows a digital microscopy image of the capillary tube. The measured diameter is 132.7±0.9 µm. The relatively larger standard deviation from the computer algorithm compared to digital microscopy is due to: 1) the OCT imaging of tube phantom (containing the scattering media) has lower contrast compared to digital microscopy imaging of tube in air; 2) OCT imaging system has lower resolution (10 µm) compared to that of digital microscopy (~1 µm). Therefore, the tube edge in OCT image is not as sharp as those in the digital microscopy, which will result in errors in segmentation. Nevertheless, the result shows that the mean of estimation obtained by the automatic computer analysis is comparable to the true tube dimension.
3.2 OCT imaging and quantification of human kidney structures
The kidney microstructures of interest, including the uriniferous tubules, vessels, and glomeruli were identified based on their distinct morphologies. Comparisons between the OCT image and the corresponding histological micrograph indicated a close match in terms of the main structural features. In addition, the resolution of the OCT images (~10 µm) was sufficient for the purpose of revealing the morphological details. Figure 6(a) shows a representative cross-sectional OCT image of the human kidney, and Fig. 6(b) is the corresponding histopathology. As seen in Fig. 6(a), tissues with high backscattering such as kidney capsule appeared bright, while low backscattering regions such as uriniferous tubular lumens appeared dark. It was clearly observable that OCT could penetrate through the kidney capsule (C) with a penetration depth of more than 800 µm. The kidney microanatomy including uriniferous tubules (T) and glomeruli (G) were also readily distinguished.
Figure 6(c) shows the histogram of the tubular lumen diameter measured by the automatic algorithm described in Section 2.3. Automatic measurement gives an estimation of lumen diameter of 27.5±10.1 µm. The manual measurement of lumen diameter from the histology slide (Fig. 5(b)) gives the results of 29.5±9.2 µm (see Fig. 6(d)). This shows that the results obtained by the automatic computer analysis are comparable to that of the manual measurements of histology slide. However, the computer calculation was much faster than the manual measurements. In addition, computer-aided analysis promises to automatically analyze a large volume of data (for example, three-dimensional data) efficiently and will be very helpful for providing the clinicians with quantitative information in a timely manner.
3.3 Three-dimensional imaging visualization of human kidney
3.3.1 Human kidney case I (blood vessels)
Figure 7(a) is the three-dimensional view of the human kidney, as generated from individual cross-sectional images. Figures 7(b)–(d) shows representative images along the three orthogonal planes (XY, YZ, and XZ), respectively. Detailed kidney vascular networks were visualized in all the image planes. The OCT image data set was further segmented and analyzed to quantify the luminal diameter of the blood vessels. Figure 7(e) shows the 3D reconstructed images showing vascular trees after intensity segmentation. The segmented 3D vascular tree was reconstructed by utilizing a 3D visualization software (Amira). The morphological features of the blood vessels can be examined. Figure 7(f) shows the quantification of the representative blood vessels luminal diameters from the ROI. Figure 7(g) shows the volume histogram of the diameter (which is obtained by counting the voxel numbers associated with the specific diameter, and multiplied by the individual voxel volume, 150.9 µm3), indicating that the majority of vessel diameters range from 50 µm to 100 µm.
For those regions without any microstructures such as tubules or vessels, the light intensity decreases exponentially with depth because of light scattering effects. However, hollow microstructures such as uriniferous tubules or blood vessels alter this exponential decay pattern due to the minimal light scattering within these hollow structures. After the light passes through these structures, it continues decreasing again. This phenomenon results in relatively higher light intensity (hyperdense shadow) below some of the microstructures as shown in the cross-sectional images (Fig. 6(c) and (d)), and casts white shadows on the en face image (Fig. 6(b)).
3.3.2 Human kidney case II (uriniferous tubules)
The previously described procedures were applied to another kidney as shown in Fig. 8. As with case I, detailed kidney tubular structures were visualized in all image planes (Fig. 8(a)–(d)). Figure 8(e) shows the 3D reconstructed images of the tubular network after intensity segmentation, which allows comprehensive examination of morphological features and interconnectivity of the renal tubules. Figure 8(f) shows the automatic quantification of tubular diameters, which were color-coded on the structural map. The volume histogram in Fig. 8(g) indicates that most tubule luminal diameters at this region range from 20 µm to approximate 40 µm, with a mean diameter around 30 µm.
3.3.3 Human kidney case III (distended uriniferous tubules)
Case III is from a third kidney. Figures 9(a)–(d) show 3D cut-through views and the representative images along the three orthogonal planes. Two clusters of distended tubules are clearly identified on Fig. 9(e). Figure 9(f) shows the automatic quantification of all the luminal tubular diameters. The tubular lumen diameters range approximately from 30–60 µm in diameter, as shown in the volume histogram (Fig. 9(g)).
3.3.4 Human kidney case IV (glomerulus)
We were also able to visualize the glomerular structures. Figures 10(a)–(d) shows an individual glomerulus in the enlarged view, as compared to previous figures. We could easily visualize the glomerulus surrounded by the circular Bowman’s space. However, the segmentation of the complete glomeruli was challenging, because in most cases, the Bowman’s space separating the glomerular capillary tufts from the renal parenchyma is not a full circle (see Fig. 10(d) and Fig. 6(b)). Figure 10(d) shows the diameter of the glomerulus to be approximately 220 µm. This result is in agreement with previous literature using ultrasound imaging (216±27 µm) . Figure 10(e) shows the 3D view of the segmented Bowman’s space.
3.3.5 Human kidney case V (vessels, tubules, & glomeruli)
Figure 11 shows a representative region with different renal structures including blood vessels, uriniferous tubules, and glomeruli. The glomeruli are surrounded by an expanded network of uriniferous tubules and blood vessels. The diameters of the glomeruli are approximately 200 µm. This result demonstrates the capability of OCT to visualize different renal microstructures in situ.
OCT is a rapidly developing imaging modality that can produce 3D imaging of tissue in situ and in real time. OCT can provide cross-sectional images which make 3D reconstruction and image processing possible. It can visualize tissue microstructure without the need for contact or tissue removal, thereby facilitating sterility and minimizing possible damage to the tissue.
OCT’s high resolution capability is sufficient for imaging numerous organs, for example, the human kidney. In this study, we were able to image up to approximately 800 µm depths in the human kidney, which was deep enough to image superficial blood vessels, uriniferous tubules, and glomeruli. Such structures are closely related to many physiological functions, for example, in the case of evaluation of transplant kidney function . High acquisition speed (video rate) achieved by Fourier domain OCT enables real time imaging in 3D [25–28] and the surveying a large kidney surface area in a timeframe reasonable for clinical practice. With further development, OCT has the potential to be translated into clinical settings for kidney imaging.
Prior studies utilized non-human kidneys for OCT imaging analysis, while this study uses OCT to study the human kidney ex vivo. Various structures from different human kidneys were readily distinguished, including the blood vessels, uriniferous tubules, glomeruli, and kidney capsules. The methods of this study could be directly applied to donor kidney viability analysis, since the previous study indicated that proximal tubular structure and post-transplantation renal function are closely correlated . In addition, there are existing correlations between glomerular morphology and renal diseases, i.e. mesangial proliferative glomerulonephritis , focal segmental glomerulosclerosis , Type I diabetes mellitus , and renal ischemia . Therefore, OCT’s ability to distinguish glomeruli structures is a potentially valuable tool for the diagnosis of glomerular diseases as well.
The algorithm applied in this study is an automatic imaging processing method, which could be generalized to many other imaging analyses. The dimensional calculation of a capillary tube phantom proved that the algorithm could successfully estimate actual luminal volumes. 3D visualization and volumetric rendering provided quantitative evaluations of the dimensional changes in tubular or vessel lumens. These renderings and the use of color coded 3D images can potentially provide clinicians with useful diagnostic tools. The next step of study will be to apply this method to image different donor kidney structures, and examine the correlation between the dimensions of their imaged features and post-transplantation renal function.
We should mention that there are also limitations and challenges of applying this method in kidney imaging. In the first step (automatic segmentation), speckle from OCT images could introduce artifacts. Further improvement of imaging resolution promises to reduce the speckle sizes and would improve the algorithm performance. In addition, an intensity-based segmentation algorithm was used in this study, which is subject to the setting of threshold values. In the future, more advanced segmentation algorithms, such as marker-controlled watershed segmentation , will be investigated as well.
The automatic selection of individual ROI for further analysis (step 2) represents a unique merit of this algorithm. This approach simulates human behavior using an automatic computer algorithm. After automatic selection, various morphormetric analyses could be applied to the selected ROI, including the measurement of area, diameter, and curvature. In this study, we focused on the local diameter, because it is closely related to the kidney viability .
The quantification of tubular (or vessel) diameters (step 3) was achieved by automatic identification of the boundary and skeleton of individual ROIs. This approach was limited in its estimation of the correct tubular diameter when the imaging plane did not cut through the central axis of the tubules. However, this limitation was also shared by most, if not all, cross-sectional imaging methods. For example, histology analysis of tubular and glomeruli diameter will be subject to the same sampling limitations. In our study, 3D OCT images with two orthogonal cross-sections (XZ and YZ) were utilized to obtain an averaged estimation of the tubular dimension. The sampling limitation will be further alleviated by panning the imaging plane 180 degree to fully cover the different angles. In addition, the skeleton extraction method used in this study is based on morphological thinning, which sometimes led to unwanted branches , which tended to under-estimate the tubular diameter. However, this limitation was alleviated when a large number of boundary pixels were evaluated and presented statistically. Reasonably accurate estimations of diameters of capillary tube phantom and kidney tubules were achieved through this algorithm, as confirmed by digital microscopy and histology. To overcome these above-mentioned limitations completely, future development of 3D boundary and skeleton recognition algorithms would be an ultimate solution.
We should note as well that the present image analysis method cannot automatically differentiate different kidney structures. At the present study, we focused on the automatic ROI selection and quantification of morphometric parameters. The next step will be to develop an automatic classification algorithm to differentiate different structures such as tubules, blood vessels, and glomeruli. This work will involve new algorithms such as pattern recognition for automatic classification.
Although the present OCT system has a limited resolution of 10 µm, it is still sufficient to detect the tubules in human kidney. We observed tubular diameters range from 30–60 µm from four human kidneys after formalin fixation. From the literature, normal human proximal convoluted tubule has a diameter ~55 µm . The ultimate clinical utility of this method will be assessed by the clinical evaluation of kidney viability, where the threshold tubular diameter for viable kidney can be determined.
In summary, OCT imaging of human kidney was visualized in real time and an automatic image analysis algorithm has been developed for quantifying spatially-resolved tubular diameter as a biomarker for kidney viability. Images along the three orthogonal image-planes (XY, YZ, and XZ) in the Euclidean space were displayed sequentially. Moreover, the rendering of the images provided a 3D volumetric view. The computed microstructure sizes were then color-coded on the reconstructed images, revealing quantitative information of the kidney microanatomy. Based on the results of this study, we have demonstrated the capability of OCT imaging and automatic quantification of human kidney microanatomy. The ability of OCT to provide 3D, high resolution imaging illustrates the potential of using OCT to image donor kidney structures and to evaluate the organ’s viability, or image the responses to acute kidney injuries. Future work will involve the quantification of those parameters for different human kidneys to obtain the baseline values for diagnostic purposes, and perform OCT images for human kidney in vivo to further analyze and diagnose kidney diseases.
We thank Anik Duttaroy, Bobak Shirmahamoodi, Dennis Truong, and Dipankar Dutta for technical assistances. This work is supported in part by the Nano-Biotechnology Award of the State of Maryland, the Minta Martin Foundation, the General Research Board (GRB) Award of the University of Maryland, the University of Maryland Baltimore (UMB) and College Park (UMCP) Seed Grant Program, the Prevent Cancer Foundation, and the National Kidney Foundation of the National Capital Area.
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, 1178–1181 (1991).
2. M. R. Hee, J. A. Izatt, E. A. Swanson, D. Huang, J. S. Schuman, C. P. Lin, C. A. Puliafito, and J. G. Fujimoto, “Optical coherence tomography of the human retina,” Archives of Ophthalmology 113, 325–332 (1995).
3. W. Drexler, H. Sattmann, B. Hermann, T. H. Ko, M. Stur, A. Unterhuber, C. Scholda, O. Findl, M. Wirtitsch, J. G. Fujimoto, and A. F. Fercher, “Enhanced visualization of macular pathology with the use of ultrahigh-resolution optical coherence tomography,” Archives of Ophthalmology 121, 695–706 (2003).
4. M. Wojtkowski, V. Srinivasan, J. G. Fujimoto, T. Ko, J. S. Schuman, A. Kowalczyk, and J. S. Duker, “Three-dimensional retinal imaging with high-speed ultrahigh-resolution optical coherence tomography,” Ophthalmol. 112, 1734–1746 (2005).
5. J. G. Fujimoto, S. A. Boppart, G. J. Tearney, B. E. Bouma, C. Pitris, and M. E. Brezinski, “High resolution in vivo intra-arterial imaging with optical coherence tomography,” Heart 82, 128–133 (1999).
6. I. K. Jang, B. E. Bouma, D. H. Kang, S. J. Park, S. W. Park, K. B. Seung, K. B. Choi, M. Shishkov, K. Schlendorf, E. Pomerantsev, S. L. Houser, H. T. Aretz, and G. J. Tearney, “Visualization of coronary atherosclerotic plaques in patients using optical coherence tomography: comparison with intravascular ultrasound,” J. Am. College Cardiology 39, 604–609 (2002).
7. B. E. Bouma, G. J. Tearney, C. C. Compton, and N. S. Nishioka, “High-resolution imaging of the human esophagus and stomach in vivo using optical coherence tomography,” Gastrointestinal Endoscopy 51, 467–474 (2000).
8. M. V. Sivak Jr., K. Kobayashi, J. A. Izatt, A. M. Rollins, R. Ung-Runyawee, A. Chak, R. C. Wong, G. A. Isenberg, and J. Willis, “High-resolution endoscopic imaging of the GI tract using optical coherence tomography,” Gastrointestinal Endoscopy 51, 474–479 (2000).
9. X. D. Li, S. A. Boppart, J. Van Dam, H. Mashimo, M. Mutinga, W. Drexler, M. Klein, C. Pitris, M. L. Krinsky, M. E. Brezinski, and J. G. Fujimoto, “Optical coherence tomography: advanced technology for the endoscopic imaging of Barrett’s esophagus,” Endoscopy 32, 921–930 (2000).
10. Y. Chen, A. D. Aguirre, P. L. Hsiung, S. Desai, P. R. Herz, M. Pedrosa, Q. Huang, M. Figueiredo, S. W. Huang, A. Koski, J. M. Schmitt, J. G. Fujimoto, and H. Mashimo, “Ultrahigh resolution optical coherence tomography of Barrett’s esophagus: preliminary descriptive clinical study correlating images with histology,” Endoscopy 39, 599–605 (2007).
11. J. G. Fujimoto, “Optical coherence tomography for ultrahigh resolution in vivo imaging,” Nature Biotechnol. 21, 1361–1367 (2003).
12. K. W. Gossage, T. S. Tkaczyk, J. J. Rodriguez, and J. K. Barton, “Texture analysis of optical coherence tomography images: feasibility for tissue classification,” J. Biomed. Opt. 8, 570–575 (2003).
13. X. Qi, M. V. Sivak, G. Isenberg, J. E. Willis, and A. M. Rollins, “Computer-aided diagnosis of dysplasia in Barrett’s esophagus using endoscopic optical coherence tomography,” J. Biomed. Opt. 11, 044010 (2006).
14. 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, 054055 (2008).
15. Y. Chen, A. D. Aguirre, P. Hsiung, S. W. Huang, H. Mashimo, J. M. Schmitt, and J. G. Fujimoto, “Effects of Axial Resolution Improvement on Optical Coherence Tomography (OCT) Imaging of Gastrointestinal Tissues,” Opt. Express 16, 2469–2485 (2008).
16. Y. Chen, P. M. Andrews, A. D. Aguirre, J. M. Schmitt, and J. G. Fujimoto, “High-resolution three-dimensional optical coherence tomography imaging of kidney microanatomy ex vivo,” J. Biomed. Opt. 12, 034008 (2007).
18. P. M. Andrews, Y. Chen, M. L. Onozato, S. W. Huang, D. C. Adler, R. A. Huber, J. Jiang, S. E. Barry, A. E. Cable, and J. G. Fujimoto, “High-resolution optical coherence tomography imaging of the living kidney,” Lab. Investigation 88, 441–449 (2008).
19. O. B. Franc, C. Stefano, V. S. Maria, G. Lorenzo, H. Yale, T. Stefano, and S. Antonio, “Automatic evaluation of arterial diameter variation from vascular echographic images,” Ultrasound Med. Biol. 27, 1621–1629 (2001).
20. V. Gemignani, F. Faita, L. Ghiadoni, E. Poggianti, and M. Demi, “A system for real-time measurement of the brachial artery diameter in B-mode ultrasound images,” IEEE transactions on Medical Imaging 26, 393–404 (2007).
21. R. Huber, M. Wojtkowski, and J. G. Fujimoto, “Fourier Domain Mode Locking (FDML): A new laser operating regime and applications for optical coherence tomography,” Opt. Express 14, 3225–3237 (2006).
22. R. C. Gonzalez and R. E. Woods, Digital Image Processing (Prentice Hall, 2007).
23. L. Lam, S.-W. Lee, and C. Y. Suen, “Thinning Methodologies-A Comprehensive Survey,” IEEE Trans. Pattern Anal. Machine Intelligence 14, 879 (1992).
25. N. A. Nassif, B. Cense, B. H. Park, M. C. Pierce, S. H. Yun, B. E. Bouma, G. J. Tearney, T. C. Chen, and J. F. de Boer, “In vivo high-resolution video-rate spectral-domain optical coherence tomography of the human retina and optic nerve,” Opt. Express 12, 367–376 (2004).
26. R. A. Leitgeb, W. Drexler, A. Unterhuber, B. Hermann, T. Bajraszewski, T. Le, A. Stingl, and A. F. Fercher, “Ultrahigh resolution Fourier domain optical coherence tomography,” Opt. Express 12, 2156–2165 (2004).
27. M. Wojtkowski, V. J. Srinivasan, T. H. Ko, J. G. Fujimoto, A. Kowalevicz, and J. S. Duker, “Ultrahigh resolution, high speed, Fourier domain optical coherence tomography and methods for dispersion compensation,” Opt. Express 12, 2404–2422 (2004).
28. R. Huber, M. Wojtkowski, J. G. Fujimoto, J. Y. Jiang, and A. E. Cable, “Three-dimensional and C-mode OCT imaging with a compact, frequency swept laser source at 1300 nm,” Opt. Express 13, 10523–10538 (2005).
29. S. Daiman and I. Koni, “Glomerular enlargement in the progression of mesangial proliferative glomerulonephritis,” Clin. Nephrol. 49, 145–152 (1998).
31. H. J. Gunderson and R. Osterby, “Glomerular size and structure in diabetes mellitus II. Late abnormalities,” Disbetologia 13, 43–48 (1977).
32. K. Moran, J. Mulhall, D. Kelly, S. Sheehan, J. Dowsett, P. Dervan, and J. M. Fitzpatrick, “Morphological changes and alterations in regional intrarenal blood flow induced by graded renal ischemia,” J. Urology 148, 463–466 (1992).
33. W. F. Ganong, Review of Medical Physiology, The McGraw-Hill Companies, (2005).