1.

Introduction

Bladder cancer is among the most frequent types of malignancy and one of the leading causes of death in the United States, with estimated 61,420 new cases diagnosed and 13,060 deaths in 2006 (Ref. 1). Most bladder cancers are superficial carcinomas and originate in the basal cell layer of the urothelium. Thus, early diagnosis is crucial for appropriate treatment and enables enhancement of the cure rate and survival of patients. Yet this remains a clinical challenge. For instance, screening techniques such as cytology,² fluorescence in situ hybridization,³ and bladder tumor antigen (BTA) (Ref. 4), are very sensitive for clinical detection of carcinoma in situ (CIS) and high-grade transitional cell carcinomas (TCCs); however, these techniques are much less sensitive for the early diagnosis of low-grade TCCs and are unable to locate the cancerous lesions, which compromises their ultimate effectiveness. On the other hand, conventional medical imaging methods such as x-ray, computed tomography (CT), and magnetic resonance imaging (MRI) fail to detect early bladder malignancies because of their resolution limitation. As a result, cystoscopy remains the current clinical standard for early bladder cancer detection. However, as cystoscopy is an en face imaging modality and provides only surface imaging without displaying any morphological details below the bladder surface, it often misses flat early bladder cancers such as CIS that may be visually normal or unspecific; thus, random biopsy is necessary for conclusive diagnosis and staging of malignancy for patients with positive findings in urine cytology. However, this invasive procedure may still miss approximately 50% of early flat bladder cancers.^{5, 6} Therefore, a noninvasive and high-resolution imaging technique would be highly desirable to improve the diagnostic sensitivity and specificity, critical to the therapeutic benefits of early bladder cancer patients.

Recent advances in optical imaging have demonstrated a number of potential modalities for early bladder cancer detection. For instance, fluorescence cystoscopy following photosensitizer administration has significantly enhanced diagnostic sensitivity by detecting^{7, 8} photochemical changes pertinent to TCCs. Optical coherence tomography (OCT), on the other hand, can provide high-resolution cross-sectional images of bladder morphology at a depth of $1\text{to}2\mathrm{mm}$ , which is advantageous for detection of various types of superficial lesions.^{9, 10, 11} Recent technological advances of spectral domain OCT (SDOCT) has dramatically improved the dynamic range and imaging rate.^{12, 13} With optical configurations in spectral domain that circumvent the slow mechanical scanning in the reference arm, imaging rate has been improved up to several hundredfold and potentially enables in vivo 3-D OCT imaging.^{14, 15, 16} In addition, direct access of spectral interference modulation enables Fourier-domain Doppler OCT (FDOTC) for real-time blood flow imaging,^{17, 18} ultra-high-resolution FDOCT (Refs. 17, 19) of retinal imaging, polarization-sensitive SDOCT (Refs. 20, 21), and spectral second-harmonic generation²² OCT to enhance image contrast in biological tissue. In addition to technological advances, clinical applications of SDOCT have already been performed in various tissues, including eye,¹⁵ esophagus,¹⁴ and urinary bladder,²³ demonstrating dramatic improvement on image fidelity and imaging rate.

In terms of bladder imaging, OCT has demonstrated the great potential to diagnose various bladder diseases, especially bladder cancer. ^{10, 24, 25, 26, 27, 28, 29} Implementations of OCT endoscopes using a rotary joint, piezoelectric transducer (PZT), paired-angle-rotation-scanning probe and microelectromechanical system (MEMS) micromirror have been reported^{30, 31, 32} for in vivo imaging diagnosis of urinary bladder cancers to fully take advantage of the OCT technique. Since the introduction of MEMS technology to OCT endoscopy,³¹ we have recently made substantial improvements on the design and packaging of MEMS-based OCT endoscopes. Here, we demonstrate our recent technological advances on spectral-domain endoscopic OCT (SDEOCT) that enables faster structural imaging and simultaneous blood flow measurement of bladder in vivo. We compare the image results obtained from time-domain OCT (TDOCT) and SDOCT systems. Then, we present diagnostic results to evaluate the utility and potential limitations of SDEOCT for in vivo bladder imaging diagnosis based on animal and preliminary clinical studies.

2.

Materials and Methods

2.1.

MEMS-Based SDEOCT

As depicted in Fig. 1 , a high-power broadband source (BBS) is employed, which can provide pigtailed output power of $12\mathrm{mW}$ at central wavelength of $\lambda =1320\mathrm{nm}$ and with full width half maximum spectral bandwidth of $\Delta \lambda =78\mathrm{nm}$ (resulting in a coherence length $L_C\approx 10\mu \mathrm{m}$ ). The light is delivered to a fiber optic Michelson interferometer in which light is divided equally to reference and sample arms. For visual guidance, the red light from a $670\text{-}\mathrm{nm}$ diode laser is coupled into the input arm via a 95%:5% fiber coupler.

Fig. 1

Schematic diagram of a fiber-optic SDEOCT system. The SDOCT system is connected to a MEMS-based OCT endoscope for in vivo bladder imaging study: BBS, broadband source ( ${\lambda }_0=1320\mathrm{nm}$ , $\Delta {\lambda }_{\mathrm{FWHM}}=78\mathrm{nm}$ , $P=12\mathrm{mW}$ ); LD, aiming laser diode $(\lambda =670\mathrm{nm})$ ; CM, fiber optic collimator; FPC, fiber polarization controller. For the endoscopic setup, lateral scanning range is up to $4.5\mathrm{mm}$ , and the spatial resolutions are $10\mu \mathrm{m}$ axially and $12\mu \mathrm{m}$ laterally. The single-frame dynamic range of the endoscopic OCT system reached $108\mathrm{dB}$ . A rod lens imaging catheter is placed parallel to OCT cystoscope for image guidance.

In contrast to TDOCT, which uses a grating-lens-based optical delay line to implement longitudinal scan, a stationary mirror is employed in the reference arm to match the optical path lengths between the sample and the reference arms of the fiber optic interferometer. The sample arm can be either connected to a bench-top stereoscope in which light exiting the fiber is collimated to $\varphi 4.5\mathrm{mm}$ , scanned laterally by a servomirror, and focused by an $f=40\mathrm{mm}$ achromatic lens onto biological tissue under examination. Through a modified FC/APC (fiber channel/angled physical contact) connector, it can also be connected to a MEMS-based pigtailed OCT endoscope, which is inserted into the $\varphi 4\mathrm{mm}$ instrument channel of a 22Fr commercial endoscopic sheath for in vivo animal or human study. As previously reported,³¹ the optical arrangement in the distal OCT scope enables visual guidance via a $\varphi 2.8\text{-}\mathrm{mm}$ scope with either white light or fluorescence image of the bladder surface. In the OCT endoscope, the light from the single-mode (SM) fiber is coupled by a 0.25-pitch selfoc lens to a $\varphi 1.0\text{-}\mathrm{mm}$ collimated beam, deflected by a fixed reflecting prism and a MEMS mirror, and is then focused by a laser doublet $(f10\mathrm{mm}∕\varphi 5\mathrm{mm})$ onto the bladder wall for OCT scan. The transverse light scanning in the OCT endoscope is facilitated by an Al-coated CMOS-MEMS mirror, which is actuated over $18°$ or $4.2\mathrm{mm}$ (limited by the scope size) in the transverse direction by an applied electrical voltage or current at a repetition rate up to $100\mathrm{Hz}$ .

In the detection fiber, the light beams returning from sample and reference arms are recombined and connected to a spectral radar in which light is collimated by a fiber optic achromatic lens $(f=55\mathrm{mm})$ , diffracted by a holographic grating $(d=1200{\mathrm{mm}}^{-1})$ , and then focused by an achromatic lens $(f=120\mathrm{mm})$ onto a 1024-pixel-line InGaAs photodiode array (Sensors Unlimited, New Jersey) mounted on a motored 3-D stage for high-precision focusing alignment. The detected spectral graph, including spectrally encoded interference fringes from different depth within the biological sample, is amplified and then digitized and streamlined to the hard disk of an image workstation via a multi-channel $12\text{-}\text{bit}$ analog-to-digital converter (A/D) (DAQ6111E, NI) at $5\mathrm{MHz}$ to enable 2-D imaging at nearly $8\mathrm{fps}$ (frames per second; post image processing). Because of complex data processing [e.g., spline interpolation, fast Fourier transform (FFT)] involved in converting spectral graphs to a reconstructed OCT image, instantaneous image processing and display is reduced to approximately $5\mathrm{fps}$ , but is sufficient for in vivo visualization of bladder wall. Data processing to reconstruct each depth profile or A-scan is implemented simultaneously with scanning of the lateral servo mirror, i.e., the MEMS mirror in the OCT endoscope following detection of each spectral graph until a 2-D image is completed and displayed. Similar to TDEOCT, the axial and the transverse resolutions of SDEOCT are determined by the coherence length $L_c$ and the focal spot size of the OCT scope, respectively. Based on the preceding parameters used in our setup, the axial resolution is $\sim 10\mu \mathrm{m}$ , and the transverse resolution is $\sim 12\mu \mathrm{m}$ .

In terms of SDOCT versus TDOCT, of particular interest is the ability of SDOCT to simultaneously image tissue architecture and subsurface blood flow/velocity (i.e., Doppler SDOCT) in nearly real time, which can be readily extracted from differential spectrograph data (i.e., phase difference) with minimal hardware modification.¹⁷ Doppler OCT imaging can provide very useful functional blood flow information pertinent to bladder cancer because abnormal angiogenesis underneath the bladder surface can be induced by bladder tumorigenesis. For instance, cancer cells can induce abnormal angiogenesis once the tumor size is greater than $1\text{to}2\mathrm{mm}$ in diameter,³³ and when the fresh nutrients and oxygen are rapidly depleted and new blood vessel formation is necessary to replenish them.³⁴ Thus, detecting abnormal angiogenesis within bladder wall can provide additional diagnostic information for bladder cancers, thereby improving diagnostic sensitivity and specificity. As previously reported,¹⁷ subsurface Doppler flow image can be acquired by subtraction of multiple $\left(N\right)$ A-scans at each lateral position. In addition to the amplitude term $I\left(z\right)=\mid \tilde{I}\left(z\right)\mid$ of the inverse complex Fourier transform $\tilde{I}\left(z\right)=i\mathrm{FFT}\left[I\left(k\right)\right]$ , which constitutes the structural SDOCT image, the phase term $\Phi (z,\tau )={\tan }^{-1}\{\mathrm{Im}\left[\tilde{I}\left(z\right)\right]∕\mathrm{Re}\left[\tilde{I}\left(z\right)\right]\}$ is extracted. And the phase variation between two adjacent A-scans at the same lateral position $\Delta \Phi (z,\tau )$ is used to reconstruct the Doppler flow velocity $v\left(z\right)=\Delta \Phi (z,\tau )\lambda ∕\left[4\pi \tau \cos \left(\alpha \right)\right]$ (Ref. 17), where $\tau$ is the interval between two A-scans. The excessive phase noise that imposes a major limitation to the dynamic range of Doppler OCT can be improved by implementing multiple scans and averaging³⁵ $\Delta \Phi (z,\tau )$ . The major advantages of such technique for in vivo Doppler imaging are obvious: it does not require hardware modification and imaging can be much faster to reduce the influence of motion artifacts.

3.

Results

To examine the imaging fidelity and suitability of MEMS-based OCT systems for in vivo bladder imaging, we performed experiments to compare endoscopic versus bench-top OCT systems and endoscopic SDOCT versus TDOCT systems. The axial resolution of the bench-top OCT setup that employed a $\varphi 4.5\mathrm{mm}$ beam collimator and a $f∕40\mathrm{mm}$ achromatic lens was approximately $\varphi 12\mu \mathrm{m}$ , equivalent to that of the MEMS-based endoscopic OCT catheter with a $\varphi 1\mathrm{mm}$ beam collimator and a $f∕10\mathrm{mm}$ achromatic lens. But the transverse laser steering range of OCT catheter was limited $(<4.6\mathrm{mm})$ by the scope size and was smaller than that of the bench-top setup $\left(6\mathrm{mm}\right)$ . Figure 2 shows the comparative 2-D cross-sectional OCT images of a porcine bladder taken by OCT endoscope [Fig. 2A] versus bench-top OCT stereoscope [Fig. 2B]. The morphological details of porcine bladder (a close mammalian homologue) such as urothelium (U), lamina propria (LP), and muscularis (M) in Fig. 2A were clearly delineated at fairly comparable image fidelity to those in Fig. 2B, except for a slight reduction $\left(2.87\mathrm{dB}\right)$ of image SNR due to light coupling loss in the endoscope. The result demonstrates the drastic improvement on our implementation of MEMS-based OCT catheter and justifies the utilities of in vivo bladder imaging study. Figure 3 shows the comparison between TDOCT and SDOCT endoscopic images. Both OCT images delineate the bladder morphology, e.g., U, LP, and M; however, a quantitative analysis between these two images indicates that at almost sixfold improved imaging rate, SDOCT [Fig. 3B] still shows a $2.5\text{-}\mathrm{dB}$ enhanced image SNR compared with TDOCT [Fig. 3A]. Therefore, SDOCT offers benefits in both image fidelity and imaging speed for endoscopic in vivo imaging study. More importantly, SDOCT can provide simultaneous Doppler image to map subsurface blood flow distribution. For instance, Fig. 4 shows the result of calibration study using tissue phantom in which the profile of two bidirectional flows (1% intralipid solution, ${\mu }_s\approx 2.4{\mathrm{cm}}^{-1}$ ) in a $\varphi 0.56\text{-}\mathrm{mm}$ translucent conduit can be quantitatively imaged by SDOCT. Based on the calibration result, Fig. 5 shows the result of in vivo porcine bladder study in which two minute blood flows (roughly $\varphi 150\mu \mathrm{m}$ ) in the lamina propria (about $800\mu \mathrm{m}$ underneath the bladder surface) were detectable, as indicated by the dashed circles with estimated apparent flow rates of 0.51 and $0.13\mathrm{mm}∕\mathrm{s}$ . It was observed during the experiment that the phase noise induced by bladder movement and handshaking compromised the dynamic range of in vivo blood flow imaging. Currently, we are implementing interpixel-shifted SDOCT³⁶ to increase imaging rate (while maintaining imaging depth) and thus to minimize the influence of motion artifacts on Doppler OCT in vivo.

Fig. 2

Comparison between (A) cystoscopic OCT image and (B) bench-top OCT image of a fresh porcine bladder: U, urothelium; LP, lamina propria; M, muscularis. Image sizes: (A) $2\mathrm{mm}\text{axially}\times 4.2\mathrm{mm}$ laterally; (B) $2\mathrm{mm}\text{axially}\times 6\mathrm{mm}$ laterally. Due to light coupling loss in the endoscope, there is a slight SNR reduction $(\sim 2.87\mathrm{dB})$ in image (A).

Fig. 3

Comparison between (A) cystoscopic TDOCT and (B) SDOCT bladder images: U, urothelium; LP, lamina propria; M, muscularis. Image size: $2\mathrm{mm}\text{axially}\times 4.2\mathrm{mm}$ laterally. (B) shows $2.5\mathrm{dB}$ enhanced image SNR compared with (A) at $\sim 6$ times faster frame rate.

Fig. 4

2-D Doppler OCT of bidirectional flows of intralipid solution: (A) structural image, (B) flow image, and (C) depth-dependent mean flow rate. Using parabolic curve fitting, the measured apparent volumetric velocity is $\sim 3.1\mathrm{ml}∕\mathrm{h}$ , which is consistent with experimental flow set of $3.0\mathrm{ml}∕\mathrm{h}$ (the slant angle was $20\mathrm{deg}$ and the camera frame rate $7.7\mathrm{kHz}$ ).

Fig. 5

Doppler SDEOCT of subsurface blood flow in a porcine bladder in vivo: U, urothelium; LP, lamina propria; M, muscularis. Image size: $2\mathrm{mm}\text{axially}\times 4.2\mathrm{mm}$ laterally. Two-minute blood flows (roughly $\varphi 150\mu \mathrm{m}$ ) were detected in the lamina propria (about $800\mu \mathrm{m}$ underneath the bladder surface). The estimated apparent flow rates were 0.51 and $0.13\mathrm{mm}∕\mathrm{s}$ , respectively.

Following successful in vivo porcine bladder imaging study, we initiated in vivo clinical study on symptomatic patients scheduled for operation-room examinations and treatments such as cystoscopy with possible biopsies, transurethral resection of bladder tumors (TURBT), and cystectomy. As a blind study to examine the diagnostic sensitivity and specificity of OCT cystoscopy, the OCT diagnosis was given during the cystoscopic procedure, and was later validated by the result of biopsied histology, which served as the gold standard of clinical diagnosis. Figure 6 shows the results for a normal human bladder, where Fig. 6A shows the surface image in vivo, Fig. 6B shows the endoscopic OCT image in vivo, and Fig. 6C shows the corresponding hematoxylin and eosin (H&E) stained histological section from the biopsied specimen. As we can see, the white light cystoscopic image with a field of view of approximately $\varphi 12\mathrm{mm}$ appears flat and uniform with plenty of visible subsurface blood vessels. Both endoscopic OCT and histological sections are $4.2\mathrm{mm}$ in the transverse direction and $2\mathrm{mm}$ in the vertical direction. It is obvious that the human bladder bears a close similarity to porcine bladder (Fig. 2) in terms of the anatomic architecture and optical properties such as scattering and absorption. The morphological details of normal human bladder delineated by OCT, e.g., the low-scattering urothelium, U; the high-scattering lamina propria, LP; and the upper muscularis, M correlated well with those in the histological section. For instance, the in vivo OCT measurement of urothelial thickness of $d_U=67.9\pm 5.6\mu \mathrm{m}$ matches the histological evaluation of $d_U=62.8\pm 4.1\mu \mathrm{m}$ , which appears as a homogeneous, low-scattering thin layer. In addition to normal bladder, Fig. 7 shows the result of a bladder tumor, including the transition to the adjacent normal bladder. The papillary appearance of the lesion is apparent under the cystoscopic surface RGB image as seen in Fig. 7A. The OCT in Fig. 7B diagnoses the lesion as papillary TCC and clearly provides the margin between the tumor and the adjacent normal bladder wall. Unlike normal bladder in Fig. 6, the distinct morphological structures such as U, SM, and M were drastically changed: the urothelium was no longer distinguishable from underlying laminar propria due to carcinogenesis in the urothelium and the resultant increase in scattering properties and homogeneity. Here, the OCT diagnosis of this TCC was based on identifying papillary features of the lesion such as the dark holes as highlighted by arrows in the thickened cancerous urothelium. Histology in Fig. 7C confirms the lesion as a papillary TCC (low grade I/III, stage T2). Note that although OCT can mark the boundary of the papillary lesion, as highlighted by the dashed line, it may not always be able to stage the invasion of cancer (e.g., $>\mathrm{T}2\mathrm{a}$ ) because of signal loss in the cancerous region. High attenuation of the OCT signal in cancerous lesions can be attributed by light loss due to increased homogeneity (such as papillary features) and enhanced subsurface vascularization.

Fig. 6

Cystoscopic OCT image of a normal human bladder in vivo: (A) surface image, (B) cystoscopic OCT image, and (C) H&E stained histological section: U, urothelium; LP, lamina propria; M, muscularis. Image sizes for (A) and (B) are $\varphi 12\mathrm{mm}$ and $2\mathrm{mm}\text{axially}\times 4.2\mathrm{mm}$ laterally, respectively. The red line in (A) indicates the lateral position of the OCT scan. (Color online only)

Fig. 7

Cystoscopic OCT image of a human bladder cancer in vivo: (A) surface image, (B) cystoscopic OCT image, and (C) H&E stained histological section: U, urothelium; LP, lamina propria; M, muscularis. Image sizes for (A) and (B) are $\varphi 12\mathrm{mm}$ and $2\mathrm{mm}\text{axially}\times 4.2\mathrm{mm}$ laterally, respectively. The red line in (A) indicates the OCT scanning position. The cystoscopic OCT image clearly provides the margin between the tumor and the adjacent normal bladder wall, as indicated by the dash curve. The lesion shown is papillary TCC (low grade I/III, stage T2). (Color online only)

4.

Discussion

In summary, we report the technical improvement of MEMS-based OCT cystoscope to enable high-resolution imaging of bladder morphology (e.g., urothelium, lamina propria, and upper muscularis) and function (e.g., subsurface blood flows) noninvasively, and the preliminary results of in vivo clinical diagnosis of transitional cell cancers. To examine the utility of cystoscopic OCT for in vivo imaging and diagnosis of bladder morphology, functions, and malformations (e.g., TCCs), we proposed a protocol that incorporated conventional cystoscopy, cystoscopic OCT, and excisional biopsy to compare the diagnostic results for each lesion, in which histological evaluation of the biopsied lesion specimens served as gold standard for clinical diagnoses. As the bladder surface images in Figs. 6A, 7A indicate, conventional cystoscopy is based on visual inspection of gross topological (i.e., superficial) changes of the bladder to diagnose bladder tumors. Therefore, it can efficiently locate and diagnose highly papillary TCCs as exemplified by Fig. 7. However, because of inability to resolve in depth direction, it is common that cystoscopy may fail to distinguish benign bladder inflammatory lesions from early flat TCCs such as CIS, both of which may show a normal or slightly reddish surface appearance.^{37, 38} Therefore, improved cystoscopic imaging techniques are required that will enable high-resolution visualization of morphological details below the bladder surface to diagnose early flat TCCs and to stage cancer invasion. Among several of the currently viable imaging techniques, OCT offers a great promise for noninvasive diagnosis of early bladder cancers, owing to its high spatial resolution (e.g., $<10\mu \mathrm{m}$ at current stage) and high dynamic range $(>100\mathrm{dB})$ at imaging depths of $2\text{to}3\mathrm{mm}$ . Based on previous animal bladder carcinogenesis studies, the unique cross-sectional imaging fashion of OCT—which is an analog to histopathology—enables the urologist to visualize the morphological architectures of the bladder critical to early cancer diagnosis such as the urothelium, the lamina propria, and the upper muscularis noninvasively, instantaneously, and at resolution below $10\mu \mathrm{m}$ , close to histopathology, if implemented endoscopically.

Recently, there have been a number of reports on endoscopic OCT techniques; however, the development of a workable high-fidelity endoscopic OCT suitable for in vivo clinical imaging diagnosis remains a technical challenge. The MEMS-based OCT endoscope presented in this paper is of particular interest to urological applications for the following reasons. First, the larger beam size (e.g., $⩾\varphi 1.0\mathrm{mm}$ ) provides improved transverse OCT resolution (e.g., $\sim 12\mu \mathrm{m}$ ). As we can see in Fig. 2, the cystoscopic OCT image quality is drastically enhanced to be comparable to bench-top OCT setup. Second, full-range MEMS laser scanning enables us to achieve a larger image size (e.g., $>4.5\mathrm{mm}$ ) in the transverse direction, which has been found very useful for precisely locating the cancer boundary, as shown in Fig. 7. Third, the front-view OCT catheter can be easily incorporated with surface imaging guidance for effective endoscopic OCT examination of larger areas, such as the bladder surface. This is critical, because urologist always wants to know the surface condition and the location of the lesions under examination, and the combined surface and OCT images can eliminate some false positive diagnoses by OCT such as scar or calcified lesions resulted from previous TURBT. Fourth, the unique setup that incorporates surface and OCT scopes makes it easy to focus during in vivo imaging and the OCT scope is reliable, sterilizable, and cost effective. Moreover, although the current OCT catheter is rigid, it can be easily modified to a flexible OCT catheter because only the $20\mathrm{mm}$ distal tip must be rigid. This modification may enable access to some areas of the bladder such as the bladder neck where the current rigid OCT catheter fails. As shown in Fig. 6, the morphological details of a normal human bladder are clearly delineated. In contrast to a normal bladder, the contrast and structural details in cancerous lesions, as exemplified in Fig. 7, are drastically reduced or diminished, and the effective imaging depth is decreased. Because of this, the use of endoscopic OCT for staging cancer invasion is likely to be limited to stages T1 or T2a (i.e., prior to muscle invasion). Previous systemic ex vivo rat bladder carcinogenesis studies revealed that OCT differentiated TCC from normal bladder and benign lesions such as inflammatory lesions and hyperplasia by detecting enhanced urothelial backscattering as a result of the increased nuclear-to-cytoplasmic ratio in TCCs. However, it has been found that some human TCCs may also show no obvious increase in backscattering, such as the lesions in Fig. 7. More detailed human data and analysis are required to study the relation between urothelial backscattering enhancement and cancer grading and staging. Fortunately, hyperplasia is far less common in human cases than in the animal carcinogenesis model following MNU installation, and the estimated diagnostic sensitivity and specificity based on initial human cases (20 cases so far) are over 91 and 80%, respectively. Currently, the identification of TCCs is based on abnormal changes in the urothelium such as the papillary features of TCCs, as exemplified in Fig. 7B. Based on more human data analysis and classification, potential future approaches to further improve OCT diagnosis of human TCCs in vivo will include detection of local vascularization with endoscopic Doppler OCT, quantification of urothelial backscattering for TCC grading, and endoscopic ultra-high-resolution OCT for quantification of nuclear-to-cytoplasmic ratio. The initial results of endoscopic Doppler OCT, as shown in Fig. 5, are of great interest for future detection of elevated tumor vasculature. However, we have noticed that due to phase noise induced in in vivo imaging, the dynamic range for small flow mapping is compromised. We are addressing the issue by employing common path optical topology^{39, 40} to minimize the phase drift caused by motion artifacts induced by patient or the examining physician.

5.

Conclusions

A SDEOCT imaging system was presented and its utility in bladder cancer diagnosis was examined in vivo based on porcine data and an initial clinical trial. Preliminary results demonstrate the potential of MEMS-based SDEOCT to provide the morphological details of porcine and human bladders such as the urothelium, the lamina propria, and the upper muscularis at a resolution and contrast comparable to TDEOCT and even the bench-top OCT system. The SDEOCT imaging system can provide cross-sectional OCT images of $2\mathrm{mm}$ in the depth direction and $4.5\mathrm{mm}$ in the transverse direction per scan at a frame rate of approximately $8\mathrm{fps}$ . The imaging system is fully automatic and incorporated with surface cystoscope for visual guidance and complimentary diagnosis with OCT scope. Additional in vivo results of blood flows in the lamina propria of bladder and clinical diagnosis of human TCCs were presented, suggesting that MEMS-based SDEOCT is a very promising optical biopsy technique for noninvasive and instantaneous imaging diagnosis of bladder diseases such as early diagnosis of bladder cancers.