Color Doppler optical coherence tomography (CDOCT) is a noninvasive technique for simultaneous high spatial resolution (~20 μm) imaging and high velocity resolution (~500 μm/s) imaging flowmetry in living tissues. In this paper, we demonstrate a reconstruction method which overcomes fundamental limitations on Doppler flow mapping associated with both high- and low-speed imaging. This algorithm is successful in retaining the high velocity resolution of CDOCT while eliminating motion artifact caused by slow image acquisition in samples which exhibit repetitive motion. We demonstrate reconstruction of blood flow throughout a beating Xenopus laevis heart and surrounding vasculature using gated CDOCT reconstruction.
© Optical Society of America
Recent advances in optical biomedical imaging have resulted in novel techniques for noninvasive assessment of subsurface tissue microstructure with high spatial resolution. Optical coherence tomography (OCT) uses low-coherence interferometry to image internal tissue structures with micron-scale resolution . OCT has been applied in vivo and in vitro in ophthalmic [e.g., 2,3], gastrointestinal , and dermatological  imaging studies. OCT has become viable as a clinical diagnostic tool with the recent advent of high-power, low-coherence sources and near real-time image scanning technologies [6,7]. The high resolution (<20 μm in three dimensions) and dynamic range (>110 dB) of OCT allow for in situ tissue imaging approaching the resolution of excisional biopsy.
OCT has recently been implemented for developmental biology studies in several standard animal models, including Xenopus laevis (the South African clawed frog) [9,10], Brachydanio rerio (the zebrafish) , and Rana pipiens (the American leopard frog) . Physiological and anatomical studies of the development of these species have previously been performed extensively using conventional microscopy and histology [e.g., 11,12 and references]. In vivo OCT imaging has been performed successfully on Xenopus larva, including quantitative and dynamic analysis of the developing cardiovascular  and neural  systems. Xenopus larva are particularly favorable for optical cardiac imaging due to the lack of ventral cutaneous pigment up to 35-40 days.
A recent novel extension of OCT technology is color Doppler optical coherence tomography (CDOCT), which performs micron-scale resolution velocity flow mapping simultaneous with anatomical imaging [13,14]. CDOCT has previously been demonstrated for quantitative flow analysis in the microvasculature of the hamster , chick chorioallantoic membrane , and rat dorsal skin flap . CDOCT is particularly useful as a novel contrast mechanism for identifying microscopic blood vessels, whose appearance is otherwise indistinguishable from surrounding tissue in OCT imaging in the 1300 nm wavelength region . In previous studies, quantification of flow within these vessels was achieved with velocity accuracy of less than 1 mm/s [13,14]. However, as will be shown in Theory, the minimum resolvable velocity in CDOCT is proportional to image acquisition rate . Images must thus be acquired relatively slowly (»1 s/image) in order to obtain high velocity resolution (<1 mm/s). This prevents imaging of dynamic structures due to motion artifact in living subjects. Thus the problem arises of overcoming motion artifact while retaining the capability of measuring flow in the microcirculation. In this article, we present a technique for motion-artifact-free reconstruction of flow in the beating Xenopus heart. This technique is illustrated with reconstructed CDOCT movies displaying the dynamics of the cardiovascular system simultaneous with high velocity resolution Doppler flow mapping.
The principles of OCT  and CDOCT [13,14] have been described previously in detail and will be summarized here. To obtain a map of highly localized reflection sites within tissue, OCT uses a fiber-optically integrated scanning low-coherence Michelson interferometer (Fig. 1). The coherence length of the broadband source determines the axial resolution, whereas the beam spot size sets the lateral resolution. Axial ranging measurements (i.e., A-scans) are performed by scanning the reference arm length while synchronously recording the envelope of the detected interferometric signal. Two-dimensional cross-sectional images are formed by laterally scanning the probe beam across the sample while A-scans are being acquired. Since the interferometer functions as an optical heterodyne detector, quantum noise limited detection of light backscattered by the sample may be achieved.
Doppler flow imaging using OCT is performed by using coherent detection to monitor moving scatterers within the sample . The interferometric fringe frequency detected in OCT arises from the net sum of Doppler shifts generated by the moving reference mirror (fr ) and (potentially) moving scatterers in the sample (fs ). The interferometric detector current, id ̃(t), generated by a moving scatterer in the sample is given by:
where A(t) is the amplitude of the reflectivity as a function of depth (time), and ϕ(t) is a phase term dependent on the exact axial position of the scatterer and its intrinsic backscatter spectrum . Coherent (phase-sensitive) demodulation of the detector current at the Doppler frequency induced by the constant motion of the reference arm results in the complex envelope of the interferogram:
Since each depth or A-scan is generated by movement of the reference arm and is thus a time-domain signal, and Doppler shifts (i.e., spectral information) change with depth, highly localized flow measurements are performed using joint time-frequency analysis. Namely, the short-time Fourier transform (STFT) is applied to the net detector current (comprising the summation of Eq. (2) over all moving scatterers) for each depth scan, resulting in power spectra corresponding to several “short-time” sections of the A-scan. The local Doppler frequency generated by moving scatterers is estimated from the centroid of each spectrum and related to the mean velocity, Vs , of the scatterers by:
where nt is the mean tissue index of refraction and θ is the angle between the incident beam and direction of motion of scatterers within the sample. Finally, the detected velocity is color coded to indicate the magnitude and direction of flow. These color flow images are then thresholded to eliminate velocity noise and overlaid on the corresponding OCT images for the color Doppler OCT display.
The velocity resolution, defined as the minimum resolvable velocity, , is directly proportional to the minimum detectable Doppler shift given by = 1/Nts , which is determined by the STFT window size N and the sampling increment, ts . Substituting into Eq. (3) yields :
However, the sampling increment is dependent upon the number of pixels, L, and the depth of each A-scan, D, as well as the velocity of the reference arm, vr . Relating the velocity resolution to these variables yields:
Eq. (5) states that the velocity of the reference arm , which is related to image acquisition rate, limits the velocity resolution. Finally, the image acquisition rate, or frame rate, Rf = vrρ/KD, where ρ is the axial scanning duty cycle, and K is the number of A-scans per image. The velocity resolution can be expressed compactly in terms of the physical and analysis parameters as:
Hence, for a given set of analysis parameters, Eq. (6) suggests a compromise between the desired frame rate and the minimum detectable velocity.
Our system incorporated a superluminescent diode (SLD) with a 35 nm FWHM bandwidth centered around 1270 nm, corresponding to a coherence length of 20 μm FWHM. The dynamic range for the given parameters was measured at 109 dB with a 580 μW SLD power incident on the interferometer and a 5 kHz detection bandwidth. OCT images were acquired with a reference arm velocity of 65.0 mm/s, resulting in a theoretical velocity resolution of 0.46 mm/s, assuming 16-element STFT windows. The A-scan acquisition rate was 8.0 A-scans per second.
In vivo CDOCT measurements were performed in an experimental Xenopus Laevis model. All procedures were performed in accordance with Case Western Reserve University Institutional Animal Care and Use Committee approved protocols. Stage 51  Xenopus tadpoles were immersed in 0.0125% Tricaine until they no longer responded to touch. With the sample beam incident on the ventral side of each specimen, oblique (45°) sagittal sections of the beating heart were acquired with OCT and analyzed for flow.
The heartbeat of the specimen was measured under a microscope and using OCT “optical cardiograms” . Here, the sample beam was held stationary over the ventricle to obtain reflectivity profiles as a function of time, analogous to M-mode ultrasound. The heart rate was estimated from the rate of periodic changes in the dimension of the ventricle. Reconstruction of the beating heart was performed by obtaining a 1000 A-scan OCT image which was oversampled in the lateral direction, ensuring that at least 5 A-scans were acquired per heartbeat while the sample was translated laterally by one focussed sample probe beam spot size (14 μm). From this image, separate time-gated cardiac image frames were extracted. Each of the frames was composed of A-scans occurring at the same segment of the cardiac cycle. Therefore, if the number of A-scans per beat was T A-scans, sequential frames were composed of 1000/T lateral pixels each. Gating of the image data according to the heartbeat was performed by estimating the value of T retrospectively, by selecting that value which completely eliminated motion artifact in the reconstructed frames. It must be noted that no a priori information was needed regarding the heart rate, although heart beating periodicity was required for the duration of image acquisition.
Doppler flow processing was performed off-line, requiring <10 s of computation on a 266-MHz PC for each of the extracted 200 A-scan by 256 lateral pixel frames, requiring 48,000 16-point STFT windows per frame.
4. Results and Discussion
An oblique sagittal OCT image through the ventral surface of the specimen is shown in Fig. 2. This image is composed of 1000 lateral and 256 axial pixels, spanning 2.0 mm across and 1.07 mm deep. In Fig. 2 the abscissa has been labeled with both time of acquisition and lateral distance. The relatively large liver of the specimen lies to the left of the image, neighbored antero-ventrally by the stomach. Also, outside the pericardium, a group of branched vessels is visible. Although these stationary structures are clearly delineated in the OCT image, the motion artifact, indicated by alternating dark and light vertical bands in the center of the image, blurs the image data within the pericardium. The general shape of the heart is apparent, but large structures such as the distinct chambers are not resolvable. In addition, the diaphragm, which also moves with the beating of the heart, is completely hidden.
Fig. 3 isolates a selected smaller time segment from the OCT image in Fig. 2, demonstrating the repetitive expansion and contraction of the heart that result in motion artifact. This figure illustrates that each heart beat (measured between consecutive end diastolic dimensions) comprised approximately five A-scans. Therefore, approximately every 5th A-scan occurred at the same segment of the cardiac cycle, during which the sample beam was translated laterally 10 μm, less than one focussed sample probe beam spot size. This reconstruction method effectively imitates heart beat gated image acquisition without requiring the insertion of electrodes or otherwise independent monitoring of heart rate.
Fig. 4 is a gated CDOCT reconstruction of the beating Xenopus heart (playback at 0.75 times real-time) performed using the five sequential time frames extracted from Fig. 2. Gating of the data was performed according to the procedure described in Methods, in which it was determined that heart rate was 1.6 beats per second and T=5.00±0.01. Color Doppler flow processing has been performed solely on the region of interest indicated by the rectangle enclosing the counter-propagating vessels. The bi-directional nature of the flow is demonstrated by assigning a color map to the Doppler shifts. Here, the large vessel with a positive pulsatile Doppler shift (indicated by red) is the truncus arteriosus, which is the portion of the aorta immediately after exiting the ventricle. The processed vessel outside the pericardium, generating a less pulsatile negative Doppler shift (indicated by blue), appears to be a vein. The appearance of flow in each vessel during the cardiac cycle correlates to its corresponding role in systole or diastole. The contraction of the heart during systole is immediately followed by maximum pulsatile flow in the truncus arteriosus. Also, flow into the heart through the vein occurs preceding and during expansion of the ventricle. Due to its pulsatile nature, flow within the truncus arteriosus appears briefly; however, flow through the vein is detected for a longer duration, since blood flow is more damped when returning to the heart.
Motion of the vessel walls also results in Doppler shifts in the backscattered light in addition to flow within the truncus arteriosus. In Fig. 4, the upper region of the arterial wall expands during systole, generating negative (blue) Doppler shifts that can be misinterpreted as flow. This artifactual flow, referred to as clutter , is also prevalent in clinical Doppler ultrasonography.
Fig. 5 is a magnified reconstruction of the beating heart in which Doppler processing has been performed on the entire image. In this case, the image section was translated medially and provides visualization of a bifurcation in the truncus arteriosus. Pulsatile blood flow is visible in the rightmost branch of the vessel. Blood flow is not observed in the leftmost branch, perhaps due to the incident angle of the beam on the sample. Large Doppler shifts are also apparent in the region of the ventricle. It is clear that the majority of Doppler shifts occur from motion of the ventricle itself rather than from the flow of blood within, although some periodic flow is visible within the ventricle. Turbulent flow inside the ventricle results in uncorrelated random shifts of radiation backscattered from that region, eliminating the ability to accurately detect flow. Hence, although motion artifact in the amplitude (OCT) image has been avoided, velocity artifacts are inevitable due to Doppler shifted backscatter from dynamic structures.
CDOCT is capable of identifying flow in microscopic vessels in vivo, though slow acquisition times prevent imaging of dynamic structures such as vessel walls. In addition, velocity resolution is drastically reduced in high speed OCT imaging. A reconstruction of the dynamic cardiovascular system of the Xenopus has been shown by retrospective gating of the cardiac cycle. This method allows imaging of dynamic, periodic biological processes with high spatial acuity, without a priori knowledge of the rate of the processes. This technique requires no additional hardware to current CDOCT equipment. Since gating assumes that the dynamic process under analysis is periodic throughout acquisition of the image, a potential use for this reconstruction algorithm is measurement of flow in clinical situations, in which periodic, pulsatile flow is commonly encountered. Gated reconstruction diminishes motion artifact yet preserves velocity resolution in CDOCT.
The authors wish to acknowledge the contributions of Ton van Leeuwen. This study was supported by the Whitaker Foundation and the National Science Foundation (BES-9624617). S. Yazdanfar would like to acknowledge support by a National Institutes of Health training grant (5T32-GM-07535).
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). [CrossRef] [PubMed]
2. E. A. Swanson, J. A. Izatt, M. R. Hee, D. Huang, C. P. Lin, J. S. Schuman, C. A. Puliafito, and J. G. Fujimoto, “In vivo retinal imaging by optical coherence tomography,” Opt. Lett. 18, 1864–1866 (1993). [CrossRef] [PubMed]
3. J. A. Izatt, M. R. Hee, E. A. Swanson, C. P. Lin, D. Huang, J. S. Schuman, C. A. Puliafito, and J. G. Fujimoto, “Micrometer-scale resolution imaging of the anterior eye in vivo with optical coherence tomography,” Arch. Ophthalmol. 112, 1584–1589 (1994). [CrossRef] [PubMed]
4. J. A. Izatt, M. D. Kulkarni, H.-W. Wang, K. Kobayashi, and M. V. Sivak, “Optical coherence tomography and microscopy in gastrointestinal tissues,” IEEE J. Sel. Top. Quantum Electron. 2, 1017–1028 (1996). [CrossRef]
6. A. M. Sergeev, V. M. Gelikonov, G. V. Gelikonov, F. I. Feldchtein, N. D. Gladkova, and V. A. Kamensky, “Biomedical diagnostics using optical coherence tomography,” OSA Trends in Optics and Photonics on Advances in Optical Imaging and Photon Migration, R. R. Alfano and J. G. Fujimoto, eds. (Optical Society of America, Washington, DC1996) 2, 196–199.
7. G. J. Tearney, M. E. Brezinski, B. E. Bouma, S. A. Boppart, C. Pitris, J. F. Southern, and J. G. Fujimoto, “In vivo endoscopic optical biopsy with optical coherence tomography,” Science 276, 2037–2039 (1997). [CrossRef] [PubMed]
8. S. A. Boppart, M. E. Brezinski, B. E. Bouma, G. J. Tearney, and J. G. Fujimoto, “Investigation of developing embryonic morphology using optical coherence tomography,” Dev. Biol. 177, 54–64 (1996). [CrossRef] [PubMed]
9. S. A. Boppart, G. J. Tearney, B. E. Bouma, J. F. Southern, M. E. Brezinski, and J. G. Fujimoto, “Noninvasive assessment of the developing Xenopus cardiovascular system using optical coherence tomography,” Proc. Natl. Acad. Sci. USA 94, 4256–4261 (1997). [CrossRef] [PubMed]
10. S. A. Boppart, B. E. Bouma, M. E. Brezinski, G. J. Tearney, and J. G. Fujimoto, “Imaging developing neural morphology using optical coherence tomography,” J. Neurosci. Methods 70, 65–72 (1996). [CrossRef] [PubMed]
11. P. D. Nieuwkoop and J. Faber, Normal Table of Xenopus Laevis (Daudin) (Garland, New York, 1994).
12. J. M. W. Slack, “Xenopus and other amphibians” in Embryos: color atlas of development, J. Bard, ed. (Wolfe, Singapore1994).
13. J. A. Izatt, M. D. Kulkarni, S. Yazdanfar, J. K. Barton, and A. J. Welch, “In vivo bidirectional color Doppler flow imaging of picoliter blood volumes using optical coherence tomography,” Opt. Lett. 22, 1439–1441 (1997). [CrossRef]
14. Z. Chen, T. E. Milner, S. Srinivas, X. Wang, A. Malekafzali, M. J. C. van Gemert, and J. S. Nelson, “Noninvasive imaging of in vivo blood flow velocity using optical Doppler tomography,” Opt. Lett. 22, 1119–1121 (1997). [CrossRef] [PubMed]
15. J. A. Izatt, M. D. Kulkarni, K. Kobayashi, M. V. Sivak, J. K. Barton, and A. J. Welch, “Optical coherence tomography for biodiagnostics,” Opt. Photon. News 8, 41–47 (1997). [CrossRef]
16. M. D. Kulkarni and J. A. Izatt, “Spectroscopic optical coherence tomography,” Conference on Lasers and Electro-Optics , 1996 OSA Technical Digest Series (Optical Society of America, Washington, D.C.1996) 9, 59–60.
17. F. W. Kremkau, “Principles and pitfalls of real-time color flow imaging,” in Vascular Diagnosis, 4th ed., E. F. Bernstein, ed. (Mosby-Year Book, Inc., Missouri, 1993).