Optical micro-angiography (OMAG), based on Fourier domain optical coherence tomography (OCT), is a recently developed imaging modality that provides dynamic blood flow imaging within microcirculation tissue beds in vivo. This paper presents its first application in imaging the blood circulations in posterior chamber of human eye. To eliminate/minimize the motion artifacts in OMAG flow image caused by the inevitable subject movement, we describe a method to compensate the bulk tissue motion by use of phase changes in sequential OCT A scan signals. By use of a fast OMAG/OCT imaging system at ~840nm wavelength band, we show that OMAG is capable of providing volumetric vasculatural images in retina and choroids, down to capillary level imaging resolution, within ~10 s. The depth-resolved volumetric views of the separate retina and choroid vasculatures are also presented. In the end of this paper, we provide a comparison of the OMAG results with those from Doppler OCT and optical coherence angiography.
©2008 Optical Society of America
Optical coherence tomography (OCT) [1,2], which has the capability of producing depth-resolved, three-dimensional (3D) and high resolution images of biological tissue both in vitro and in vivo, has been used in increasingly more medical research and diagnostic applications since it was first reported in 1990s . In addition to imaging of microstructural features of the sample, OCT development has further evolved into phase-resolved optical Doppler tomography (PRODT) technology that is able to image blood flow in tissue by evaluating the phase information between adjacent A-line scans [4–10]. PRODT has seen some successful applications in imaging the dynamic blood flow in human retina in vivo, for example in [12–16]. In ophthalmological applications, the blood vessel network map (or angiogram) over the retina and choroids is often useful to aid the diagnosis and treatment for clinicians and physicians. Because the dynamic range of the measured phase difference between adjacent A scans is low, the angiogram based on the contrast of phase difference is often unacceptable. To solve this problem, Makita et al  proposed a method to use the squared value of the phase-difference to produce the angiogram over retina by the PRODT technique, which name was given as optical coherence angiography (OCA). For high resolution and 3D imaging, OCA has shown its potential to study the ocular circulation . However, it is difficult for OCA to provide accurate choroidal vessel network where velocity of blood flow is relatively high compared with that in retina. In addition, because the phases are strongly dependent on the strength of OCT signals , OCA method to image the blood vessels in the choroids is very noisy due to that the OCT signals in this layer are strongly attenuated. Furthermore, the phases are also related to the heterogeneous property of tissue . The phase noise due to the optical properties of tissue naturally raises a noise floor for PRODT measurement. This makes OCA difficult to image the capillary vessels, even in retina layer. These facts, nevertheless, restrict somehow the ophthalmological applications for PRODT and OCA.
In order to image the blood flows within the choroidal layer, Yasuno et al demonstrated a scattering optical coherence angiography method (S-OCA) . This method is based on the light absorption and the scattering properties of blood. The basic principle of S-OCA is to apply a segmentation method using intensity threshold-based binarization (ITB) to extract the chorodial vessels, which represent the low OCT signals and is otherwise obscured in OCA method. Although S-OCA images of blood vessel network in human choroidal layer has been presented , the interpretation of the S-OCA images is difficult, if not impossible, largely due to the fact that the contrast of S-OCA image is produced by the scattering OCT signals that renders it difficult to distinguish the blood flow signals from the structural signals. To solve this problem, an alternative solution is therefore needed to faithfully separate the blood flow signals from the structural signals.
Recently, based on the full range complex frequency domain optical coherence tomography (FDOCT) [19,20], our group has developed an imaging technique called 3D optical micro-angiography (OMAG) [21,22] that provides a powerful tool to delineate the dynamic blood flow network within tissue in many aspects superior to PRODT . OMAG uses a constant modulation frequency (CMF) applied to the interferograms formed between the reference and sample beams to separate the OCT signals from the moving elements from the static ones. The imaging contrast of blood flow relies on the amplitude information of OCT signals, thus it tolerates more the sample movement and the tissue heterogeneity. The less strict requirement of the phase stability for the experimental environment, on the other hand, increases the imaging speed. By use of OMAG imaging technique, Wang et al demonstrated in vivo imaging of cerebral vascular circulation of adult living mice, down to capillary level, with the skull [21, 23] and skin  left intact. It seems that it is straightforward to apply OMAG to image the blood circulations within retina and choroids. However, unlike the small animal imaging where the subject can be restrained to keep it relatively motion-free, for in vivo 3D imaging of human eye, the main restriction is the head and eye movement, which can cause strong bulk motion artifacts. These artifacts would lead to a reduction of the accuracy and quality of the resulted flow images. Thus, a solution to the motion artifacts would be needed for OMAG imaging in ophthalmology.
In this article, we present the first application of OMAG imaging of human retina and choroids in vivo. We describe a method to compensate the motion artifacts in OMAG in which the phase information resulted from PRODT is used to estimate the subject movement, and then the estimated phases are retrospectively used in the OMAG algorithm to stabilize the subject. We report that the 3D OMAG imaging of the blood circulations in retina and choroids, down to capillary level resolution, can be obtained within ~10 seconds. 3D OMAG images of retinal and choroidal vasculatures are presented, and then compared with those obtained from OCA and S-OCA approaches.
2. OMAG system setup
The system used to achieve OMAG imaging of retina and choroids is illustrated in Fig. 1, which is similar to that described previously . The system used a superluminecent diode with a central wavelength of 842nm, which has a measured axial resolution of ~8μm. The light was coupled into a fiber-based Michelson interferometer. The reference light was delivered to a reference mirror that was kept stationary throughout imaging. The sample light was coupled into a probe, which was used to deliver the light onto, and collect the light backscattered from the posterior part of the eye. The optical power of the beam on the cornea is ~700 μW, which is lower than the ANSI exposure limit . As illustrated in the Fig. 1, the probe is consisted of a pair of X-Y galvanometer scanner, a collimating lens, an objective lens and an ocular lens. The lateral imaging resolution was approximately 16 μm. The light returning from the reference and sample arms was recombined and sent to a custom designed high speed spectrometer, consisting of a 30 mm focal length collimator, a 1200 lines/mm diffracting grating and an achromatic focusing lens with 150 mm focal length. The spectrometer has a designed spectral resolution of 0.055nm, resulting in an imaging depth of approximately 6.4mm in air (i.e. the full depth). The signal sensitivity of 95 dB was measured at z=+0.5mm and dropped to 80 dB at z=+2.0mm when the camera integration time was set at 34.1 μs. In the reference arm, we used a 2cm water chamber to pre-compensate the wavelength dispersion effect caused by the human eye. Further dispersion in the system were compensated in the post data processing as described in .
The probe beam was scanned in the lateral direction (x axis shown in Fig.1) by X-scanner, driven by 20 Hz saw-tooth waveform with an amplitude ~2.5mm. In the elevational direction (y axis), the beam was scanned by Y-scanner which was driven at 0.1 Hz with an amplitude ~2.5mm too. The camera integrating time was set at 49 μs for imaging, allowing 1 μs for downloading the spectral data from CCD (1024 pixels, A scan) to the host computer via CameraLinkTM and a high-speed frame grabber board (PCI 1428, National Instruments, USA).
In the lateral direction, the B-scan was consisted of 1000 A-scans; and in the elevational direction, there were 200 discrete points. Thus, the final volume data cube of 1024 by 1000 by 200 (z-x-y) voxels was built from which the 3D structural and flow images were computed. It took ~10 seconds to obtain such volume data cube using the current setup. The operations for probe scanning, data acquisition, data storage and hand-shaking between them were controlled by a custom software package written in C++ language.
For OMAG imaging, it is necessary to introduce a constant frequency modulation in the spectral interferograms so that the OCT signals from the moving and static elements can be separated. There is a number of ways to achieve this function of frequency modulation, for example the reference mirror movement in a constant velocity during the OCT B mode scanning , and offsetting the sampling beam on the X scanner that was used to obtain the B scan OCT image [26–28]. In this paper, we used the latter approach to achieve the OMAG imaging because it is more cost effective and easier to setup. By offsetting the sampling beam on the scanner, we introduced the modulation frequency of 800 Hz (measured) into the spectral interferograms in this study, that corresponds to the minimum resolvable velocity of ~0.34 mm/s that can be measured by the OMAG system.
3.1 OMAG flow imaging
The OMAG imaging of blood flow has been presented in . To describe the method that follows to compensate the motion artifacts, it is necessary here to review briefly the OMAG method. Assuming the constant modulation frequency introduced in the interferograms is fM, then for simplicity, the spectral interferogram that is detected by the camera in OMAG or FDOCT can be expressed as:
where f0, and fD are the frequency components that correspond to the microstructural and flow information within sample, respectively, and φ is the random phase term. t1=2k, and k is the wave number, t2 is the timing capturing every A-line along lateral position. The analytic function of Eq. (1) against t2 can be constructed through the well known Hillbert transformation if the Bedrosian theorem holds [29,30] which states that if the modulation frequency fM - fD does not overlap the signal bandwidth caused by the random phase fluctuation term φ Under this condition, if fM – fD>0, which means large enough particle movement, v⇀s, that caused the Doppler frequency shift towards fM, , the analytic function of Eq. (1) can be written as:
where j = √-1 . Whereas, if fD – fM>0, Eq. (2) should be written as:
Then we perform the Fourier transformation against the time variable t1, the frequency component f0 of Eq. (2) is placed in the positive space in the entire Fourier plane, while it sits on the negative space for Eq. (3). The signals in the positive space are used for micro-structural image, i.e. conventional OCT image, and those in the negative space are used for blood flow image.
3.2 Compensation of motion artifacts.
The above analysis does not take into account of the sample movement, which is inevitable for in vivo experiments, particularly for imaging of human eye. The relative sample motion represents as a Doppler shift in the light frequency between the light it emits and that it receives. This shift is imparted when the light strikes an object which is moving, relative to the probe. For OMAG system, the Doppler frequency due to the sample motion is depicted as an additional modulation frequency, fO, in the interferogram. Therefore, the decisive frequency mixing in the interferogram, as outline in the last section, will now become fM – fD + fO that determines whether the Hilbert transform of Eq.(1) is turned into Eq.(2) or Eq.(3). In addition, the sample movement is directional when the OMAG system sees it and it is not predictable. As such, if fO is larger than fM, then the structural signals will appear in the flow image plane, giving severe artifacts on the blood flow measurement.
As an example to illustrate the effect of subject motion on the OMAG results, Figure 2 shows a typical in vivo B scan of the posterior part of human eye from a volunteer. Figure 2(A) is the conventional OCT image, where the important physiological layers, such as the retina and choroids, can be delineated. Figure 2(B) is the corresponding OMAG flow image where it can be seen that although the blood flow signals from retina and choroidal layers are evident, the subject movement deteriorates the blood flow image, particularly in the right part of the image. Further analysis from Fig. 2(B) implies that the motion is not uniform, even during one single B scan (within 50 milliseconds). This non-uniform subject movement can also be seen in the PRODT image as shown in Figs. 2(C) and 2(D).
A casual inspection of Fig. 2 shows that the blood flow imaging is better delivered by OMAG than that by PRODT, because the blood flow signals as seen in the choroidal layer in Fig. 2(B) are not seen in PDOCT images (Figs. 2(C) and (D)), as illustrated by the arrows. However, still, the subject movement artifacts are present. Thus, for in vivo OMAG imaging of human eye, a solution to the bulk motion artifacts is required so that the quality of OMAG blood flow image is improved.
Because OMAG imaging has its origin in FDOCT, the OCT signals can be directly calculated from the captured interferograms for sequential A-scans using conventional FDOCT algorithms. It is known that the OCT signal so computed is a complex function, B̃OCT (t,z), which can be written as,
where A(t,z) and ϕ(t,z) are the amplitude and the phase of B̃OCT(t,z), respectively. t is the timing when the probe beam proceeds in the B scan, and z represents the coordinate along the imaging depth. In our OMAG system, the phase term ϕ(t,z) can be expressed by the following equation:
where ϕP(t,Z) is the phase term caused by the moving red blood cells; ϕB(t,Z) is the phase term caused by the subject movement; ϕM(t,Z) is the phase term caused by the modulation frequency, fM. Following PRODT [5,6], the phase differences between the i-th and (i-1) A-lines can be written as:
where Δϕip(t,Z) and ΔϕiB(t,z) are the phase changes caused by the moving particle and the subject movement between the i-th and (i-1) A-lines, respectively. Because in OMAG system, the introduced modulation frequency,fM, is constant during imaging, ΔϕiM(t,z) is a constant value which is known a priori, i.e.
where T is the time interval between adjacent A-lines. This phase term can be directly subtracted from Eq. (6). Thus, the phase term directly related to the movement, Δϕim(t,Z) is
Because the OMAG system used in this study has an imaging speed of 20,000 A scans per second, it is reasonable to assume that the subject is static within a single A scan. Under this condition, Eq. (9) represents the bulk motion of subject between adjacent A scans, plus the phase changes due to the blood flow if presented. Usually, within a single B scan, the regions occupied by the blood vessels are much smaller than the areas by the surrounding tissue. Thus, the phase change between i-th and (i-1)-th A-scans caused by the bulk motion could be approximated by averaging all the phase differences along the A-scan,
to minimize the errors brought by the system noise and flow signals [13,31]. Here, Δϕim(X) is the bulk motion corresponding to the phase difference between i-th and (i-1)-th A-scans. Therefore, with reference to the 1st A scan in the B scan, the phases at the i-th A-line caused by the bulk subject motion, ϕim(X), can be obtained by a recursive function:
where N is the number of A scans that consists of the B scan. Thus far, the phase changes along the B scan due to the bulk tissue motion, ϕB(X,Z), have been computed. Subtract Eq. (11) from Eq. (5), and then insert into Eq.(4), we have
where the frequency modulation term due to the bulk tissue motion is eliminated. Subsequently, the spectral interferograms without the tissue motion effect can be obtained via an inverse FFT to Eq. (12),
In Eq. (10), the phase due to the bulk tissue motion is estimated by the averaged value of phase changes along the A scan. However, the phases are random for the OCT signals that fall below the system noise floor. This will have a dramatic effect on the estimation of bulk tissue motion. To reduce this effect, we used the OCT structural image to segment the regions of interest, R(x,z), within which the OCT signals are 15 dB above the system noise level. Only the phases within the region of interest are calculated. Below, we detail the steps to perform the segmentation to find R(x,z).
First, the OCT structural image was used to segment the region of interests that represents retina, and choroids. Two boundaries were determined. One corresponded to the vitreoretinal interface, i.e. anterior boundary, and the other to the deepest penetration points, posterior boundary. The anterior boundary was obtained by using an algorithm which is similar to that described by Mujat et al. . We first used a Gaussian filter with a standard deviation of 4×4 pixels to blur the cross-sectional image (B scan) obtained previously. And then the edge image was obtained by calculating the gradient of the blurred structural image. It was then converted to a binary image by setting a threshold. The first nonzero points of the binary image, representing as a curve, were detected, and the sparkle noises were eliminated. The anterior boundary was then obtained by fitting and smoothing this curve.
The posterior boundary was obtained by setting a threshold to the blurred retina structural image, which was equal to the noise floor of the obtained image. The intensity was treated as the signal if it is larger than the threshold and set it to unity; on the other hand, it was treated as noise if it is smaller than the threshold and is set to zero. Using this method, the structure image was converted to a binary image. The posterior boundary was obtained by the last nonzero points after the isolated sparkle noises were eliminated. The final posterior boundary was then obtained by fitting and smoothing procedure.
Finally, the region of region R(x,z) was obtained by bounding the anterior and posterior boundaries as described above. The phase estimation in Eq.(10) is therefore performed within this region of interest:
To further improve the accuracy of the phase estimation, the phases in Eq. (14) are only calculated for the OCT signals that are stronger than a preset threshold, that is 15dB above the system noise floor. Figure 3 illustrates the flow chart of the operations that are used to calculate the final OMAG flow images with the subject motion artifacts minimized.
To test the algorithm, we used the raw interferogram data for Fig. 2. Figure 4 show the results after the bulk tissue motion was compensated to stabilize the B scan image. The phase accumulation due to the bulk tissue motion amounted to ~250 radians within the single B scan (Fig. 4(A)). This phase changes along the B scan was used to compensate the spectral interferograms to obtain the final OMAG blood flow image as shown in Fig. 4B. Compared with Fig. 2(B), the OMAG flow image after phase compensation is almost free of motion artifacts. Such result shows that the algorithm described above works well. For comparison, we also applied the compensation method to compensate the phases in the OCT signals to obtain final PRODT flow images. The results are given in Fig. 4(C) for PRODT and Fig. 4(D) for power Doppler OCT images, respectively. Comparing between the OMAG and PRODT blood flow images, the former delivers superior imaging performance, as evidenced by 1) the blood signals circled in Fig. 4(B) is difficult to be identified in Fig. 4(C) and (D); 2) the flow signals within choroidal layer are abundant in Fig. 4(B) which are however almost absent in PRODT images.
3.3. 3D OMAG image
The 3D OMAG images were obtained by stacking the 200 B scans together that spanned 2.5mm in the elevation direction. However, a care was taken to align the misaligned B scans within C scan that was caused by the subject movement. We used the cross-correlation method, similar to that used by Shuichi Makita et al,  to align the displacement caused by the eye and head movement between adjacent frames. Because the retina has a well-layered structure, the cross-correlation function between two adjacent frames presents a sharp peak. So the displacement between them was able to be determined by the position of the peak. This method worked reasonably well for aligning the B scans within C scan to reconstruct the volumetric OMAG images.
3.4 Depth resolved retina and choroidal vasculatures images
The aligned 3D flow images are used to generate depth-resolved volumetric vasculatural images and x-y projection images. The high reflectivity layer, which is corresponding to the inner/outer segment junction (IS/OS), retina pigment epithelium (PRE), and choriocapillaries (CC), is used to segment the OCT microstructure into two masks.
The high reflectivity layer is segmented to determine the retina and choroid vessel regions. Because the high reflectivity layer is very similar to the anterior boundary of the retina, the same gradient method was applied to identify the boundary layer between retina and choroids. To do so, the axial positions of the maximum gradient 20 pixels below the anterior boundary (that was already obtained) were detected. Then all of these positions were expanded 5 pixels to both the retina side and choroidal side. After fitting and smoothing, we obtained 2 smooth curves. The upper one was used as the posterior boundary of the retina region and the lower one was used as the anterior boundary of the choroidal region.
Combining these two curves with the anterior boundary and the posterior boundary obtained in the Section 3.2, two masks were generated: one for retina layer and another for choroidal layer. These two masks were then applied to the 3D OMAG flow images obtained in Section 3.3, so that two 3D vasculatural images were arrived, one representing retinal vasculatures and another the choroidal vasculatures. A Gaussian filtering was applied to these 3D images to suppress the spurious noise and discontinuities in the vasculatures. To produce two-dimensional angiogram, the x-y projection image was produced by projecting the maximum amplitude value along the axial (z) direction in the 3D OMAG flow image on to the x-y plane (en-face projection image).
4.1 In vivo OMAG images
To demonstrate the performance of OMAG imaging of the posterior chamber of eye in vivo, we performed experiments using the system described in Section 2 on the volunteers in our laboratory. The whole experiment was done in a darkroom. To reduce the eye and head movement, the volunteer was asked to steer at a fixed position during the experiment and at the same time, the head of the volunteer was placed on a home-made stage, which helped the volunteer to loose the stress during the experiment. Every 3D spectral interferogram data set captured by OMAG system was composed of 1000 A-scan on a frame (B scan) and 200 frames in the elevation direction (C scan), from which the volumetric structural and blood flow images were obtained. This data volume was acquired over an area ~2.5 mm × 2.5 mm in approximately 10 seconds.
Figure 5 shows the in vivo imaging results produced by one volume data set captured at a position near to the optic disk. Figure 5(A) gives the OCT fundus projection image obtained by integrating signals along the depth direction from the volumetric OCT structural images using the method proposed by Jiao et al , where the major blood vessels over the retina can be seen, but not in the choroid. Figure 5(B) represents one cross-sectional image within the volumetric OCT image at the position marked by the yellow line in Fig.5(A). Figure 5(C) and the associated movie show the flying through 2D OMAG flow images of the cross-sections after the bulk motion was corrected as described in Section 3.2, where it can be seen that the blood flows within the retina and choroidal layers are clearly delineated. The results also demonstrate that the motion compensation method as described in Section 3.2 worked well for the volumetric OMAG imaging. In addition, we should pay attention to the dilation and contraction of the area of flow cross section in the movie, which are most likely caused by the heartbeat of the volunteer. In the retinal region, not only some big vessels, but also the capillary vessels are presented. Although the blood flows in the vessels in the choroid region are slow, which are difficult to be detected by PRODT, they can be captured by OMAG imaging method. These will be further discussed in Section 4.2 by comparing the OMAG images with OCA images.
Because OMAG gives the volumetric structural and flow images simultaneously, the 3D vasculatures can be merged into the 3D structural image. Such merged volumetric image is illustrated in Fig. 5(D), where a cut-through view in the center of volumetric structural image is used to better appreciate how the blood vessels innervate the tissue. To separately view the blood vessels within the retina and choroidal layers, we used the segmentation method as described in Section 3.4 to produce two masks, from the three-dimensional OMAG structural image, that represent the retina and choroids. In a cross-sectional view, the mask that represents the retina is shown as the red and yellow lines in Fig. 5(B), and that represents the choroid as the green and blue lines. To render the volumetric flow image, the blood flow signals within the boundaries of two masks are coded with green color (retina) and red color (choroid). The result is shown in Fig. 5(E), where the flow image show us two vessel networks with good connection between vessels.
To view the blood vessels in detail, Fig. 6(A) gives the x-y projection image with the blood vessels in retina coded with green color and those in choroids with red color. Fig. 6(B) and Fig. 6(C) illustrate the x-y projection images produced from blood vessels in retina and choroids, separately. Compared with Fig. 5(A), the OMAG image shows not only the retina vessels, but also the choroid vessels. What one should pay attention to is the horizontal lines visible in Fig. 6(C), the appearance of which are periodical. These lines might be caused by the heart beat of the volunteer as seen in the movie Fig. 5(C).
All of these results demonstrate that OMAG delivers superior imaging performance, not only in retina, but also in choroids, including the capability of imaging capillary vessels. These will be discussed further by comparing the images achieved by using OCA method and S-OCA method.
It should be mentioned that OMAG is capable of providing directional flow imaging of blood flows. The original demonstration of this capability was done through two 3D scans with opposite phase shift directions . This approach will inevitably slow the OMAG imaging speed. Recently, we have demonstrated a digital frequency modulation approach to achieve the directional imaging for OMAG with only one 3D scan . Further work on the directional flow imaging using OMAG is however beyond the scope of this study.
4.2 Comparison among OMAG, OCA and S-OCA imaging results
4.2.1 Comparison between OMAG and OCA
In Fig. 7, we illustrate the differences between OMAG (top row) and OCA (bottom row) images of retina (left column) and choroid (right column) vessel networks in a healthy human eye. To obtain the OCA flow image, we followed the algorithms described in  after the modulation phases (known a priori) were removed from the OCT signals. In addition to correcting the phase-wrapping errors, compensation of the bulk motion artifacts and segmentation of the regions of interests for retina and choroids described in  were applied to obtain the OCA imaging results.
For imaging the blood vessels in retina layer, OMAG shows its great advantage (comparing between Figs. 7(A) and (C)). This advantage may be appreciated by OMAG imaging of small blood vessels, marked with green ellipse in Fig. 7(A), that are however not shown up in Fig. 7(C). The advantage of OMAG imaging of the blood vessels is even more pronounced in choroidal layer when comparing between Figs. 7(B) and (D). For OCA (Fig. 7(D)), only two big vessels, marked with blue ellipses, may be distinguished from the background noise, while for OMAG (Fig. 7(B)), not only the big vessels but the capillaries are being mapped.
The reasons for causing the differences between OMAG and OCA imaging of blood vessels in retina and choroids are that OMAG imaging does not rely on the phase information of the OCT signals, while OCA is essentially a PRODT technique that uses the phases of OCT signals to measure the blood flow within tissue. Theoretically, the sensitivity of PRODT to measure the blood flow is in the order of micrometers per second, which is determined by the OCT system noise floor . However, when imaging the blood flows within tissue, the optical heterogeneity property of the sample will cause a texture background noise floor on the flow image  that reduces the system sensitivity to image the slow blood flows, particularly in the capillary vessels. In addition, in PRODT, the phases are strongly dependent on the strength of OCT signal. However, the OCT signals are attenuated in the choroidal layers which caused the PRODT image in this layer very noisy partly because phases are mathematically limited to the interval (-π, π] corresponding to the principal value of the arctangent function. These factors determine that OMAG outperforms OCA method in imaging the blood vessel networks in retina and particularly in choroidal layer. The noise characteristics may also be explained by signal to noise ratio (SNR) analysis as detailed in [35 ], where it was shown that the phase distribution may be accurately estimated only for SNR > 30dB, while the amplitude distribution is reliable for SNR > 6 dB. This implies that the OMAG method is better performed than the PRODT approach because the former is based on analysis of the amplitudes of OCT signals.
It should be mentioned that the comparison above between OMAG and OCA was made on the same data set in this study, i.e. 1000 A lines in a B scan and 200 B scan in a C scan. It is clear that if a different scanning pattern is adopted, for example denser A scan sampling in a B scan, the imaging quality of OCA would be better than that it is shown here.
4.2.2 Comparison between OMAG and S-OCA imaging in choroidal layer.
Because S-OCA is a method to image the blood vessel networks within choroidal layer, we limited our comparison of its performance with OMAG performance in imaging of blood vessels in choroids. For comparison between OMAG and S-OCA imaging of vasculatures in choroidal layer, we used the spectral interferogram data set that was used to obtain Fig. 5. Figure 8(A) and 8(B) are the resulted OMAG choroidal image and S-OCA choroidal image, respectively. To obtain S-OCA image, we followed the procedures described in , i.e. intensity threshold binarization (ITB) method, to discriminate the blood flow signals from the structural signals. We first isolated the choroidal layer from the OCT structural image and flatted it according to the highly reflective layer. After that, the mean intensity, μ, and the standard deviation, σ, of the en-face image were obtained. And then according to the structural image, the noise floor In was determined. At last the pixels that satisfy the following criteria were selected as the choroidal vessels :
where I’(x, y) is the selected intensity.
Compared to OCA choroidal image (Fig. 7(D)), S-OCA choroidal image presents a better agreement with OMAG image. Especially the two big vessels, marked with the green ellipses, are matched well with each other in OMAG and S-OCA. OMAG shows a clear vessel network with good connection between them, including capillaries. However, connection between vessels is not well presented for S-OCA, and furthermore small vessels are almost absent in S-OCA. For example, the blood vessel that marked with blue ellipse is presented in OMAG image, but it is absent in S-OCA. Further inspection between Figs. 8(A) and 8(B) shows that the appearance of the vessel networks derived from OMAG and S-OCA is quite different. It seems that some vessels in S-OCA are reversed when compared to the vessel appearance in OMAG. Further work is necessary for clarifications.
The working principle for OMAG to image the blood flow relies on the moving scatters within a scanned volume, thus the network imaged by OMAG should belong to the functional blood vessels. However, the S-OCA method is based on the structural information. The basic assumption of S-OCA method is the low scattering characteristic of choroid vessels, which assumes that the areas with low OCT signal intensity in the choroid region are occupied by the choroidal vessels. Thus, it would be difficult for S-OCA to differentiate the blood flow OCT signals from the structural OCT signals because the latter is not always of strong signal due to the signal attenuation in the choroidal layer. To further clarify this problem in S-OCA, we produced a projection image by averaging the OCT signals within the choriodal layer (choroidal mask) along the depth direction, and then the resulted image was inverted to generate a final image. In doing so, the projection image would represent the low scattering regions presented in the choroidal layer. Such structural intensity projection image for the choroidal layer is shown in Fig. 8(C). Comparing between Fig. 8(B) and Fig. 8(C), it is clear that S-OCA method selectively displays the low intensity area from the structural image, that renders S-OCA difficult to distinguish the blood vessel signals from the structural signals. The reasons for the appearance of S-OCA image may be likely due to 1) a clear tissue space that exhibits low scattering and low absorption, 2) a region that is of high absorption, and 3) a region that is obscured by a big vessel above it, i.e. blood vessel shadows. Although the artifacts caused by the blood vessel shadows may be compensated in part by a method called shadowgram , further systematic investigation on S-OCA is however beyond the scope of the current study.
In this article, we have demonstrated the potential applications of OMAG to image the ocular blood vessels in human eye in vivo. To eliminate/minimize the motion artifacts that is inevitable for in vivo imaging of human eye, we have described a method to use the phase changes between the sequential A scans within a B scan to retrospectively compensate the OCT signals in order to stabilize the B scan during the imaging. We have shown that the proposed motion compensation method worked well for in vivo volumetric OMAG imaging of the blood vessel networks in retina and choroids. We demonstrated that OMAG imaging of ocular blood vessels within both retina and choroids can be obtained within about 10 second by use of our current fast OMAG imaging system at 840nm wavelength band. With the help of the segmentation of region of interests from the volumetric structural images, the depth-resolved volumetric retinal and choroidal blood vessel networks can be visualized separately. By comparing with OCA and S-OCA images, OMAG images delivers the superior performance in imaging blood vessel networks in both the retina and choroidal layers, demonstrating the potential of OMAG imaging in ophthalmic applications.
References and links
1. A. F. Fercher, W. Drexler, C. K. Hitzenberger, and T. Lasser, “Optical Coherence Tomography - Principles and Applications,” Rep. Prog. Phys. 66, 239–303 (2003). [CrossRef]
2. P. H. Tomolins and R. K. Wang, “Theory, development and applications of optical coherence tomography” J Phys. D: Appl. Phys. 38,2519–2535 (2005). [CrossRef]
3. 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]
4. Z. P. 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]
5. Y. H. Zhao, Z. P. Chen, Z. H. Ding, H. Ren, and J. S. Nelson, “Real-time phase-resolved functional optical coherence tomography by use of optical Hilbert transformation,” Opt. Lett. 25, 98–100 (2002). [CrossRef]
6. H.W. Ren, Z. H. Ding, Y. H. Zhao, J. Miao, J. S. Nelson, and Z. P. Chen, “Phase-resolved functional optical coherence tomography: simultaneous imaging of in situ tissue structure, blood flow velocity, standard deviation, birefringence, and Stokes vectors in human skin,” Opt. Lett. 27, 1702–1704 (2002). [CrossRef]
7. R. A. Leitgeb, L. Schmetterer, C. K. Hitzenberger, A. F. Fercher, F. Berisha, M. Wojtkowski, and T. Bajraszewski, “Real-time measurement of in vitro flow by Fourier domain color Doppler optical coherence tomography,” Opt. Lett. 29, 171–173 (2004). [CrossRef] [PubMed]
10. Y. Zhao, Z. Chen, C. Saxer, S. Xiang, J. F. de Boer, and J. S. Nelson, “Phase-resolved optical coherence tomography and optical Doppler tomography for imaging blood flow in human skin with fast scanning speed and high velocity sensitivity,” Opt. Lett. 25, 114–3 (2005). [CrossRef]
12. R. A. Leitgeb, L. Schmetterer, W. Drexler, A. F. Fercher, R. J. Zawadzki, and T. Bajraszewski, “Real-time assessment of retinal blood flow with ultrafast acquisition by color Doppler Fourier domain optical coherence tomography,” Opt. Express 11, 3116–3121 (2003). [CrossRef] [PubMed]
13. B. R. White, M. C. Pierce, N. Nassif, B. Cense, B. Park, G. Tearney, B. Bouma, T. Chen, and J. de Boer, “In vivo dynamic human retinal blood flow imaging using ultra-high-speed spectral domain optical Doppler tomography,” Opt. Express 11, 3490–3497 (2003). [CrossRef] [PubMed]
14. R. A. Leitgeb, L. Schmetterer, W. Drexler, A. F. Fercher, R. J. Zawadzki, and T. Bajraszewski, “Real-time assessment of retinal blood flow with ultrafast acquisition by color Doppler Fourier domain optical coherence tomography,” Opt. Express 11, 3116–3121 (2003). [CrossRef] [PubMed]
16. R. A. Leitgeb, L. Schmetterer, C. K. Hitzenberger, A. F. Fercher, F. Berisha, M. Wojtkowski, and T. Bajraszewski, “Real-time measurement of in vitro flow by Fourier-domain color Doppler optical coherence tomography,” Opt. Lett. 29, 171–3 (2004). [CrossRef] [PubMed]
18. Y. Hong, S. Makita, and Y. Yasuno, “Three-dimensional visualization of choroidal vessels by using standard and ultra-high resolution scattering optical coherence angiography,” Opt. Express 15, 7538 (2007). [CrossRef] [PubMed]
19. R. K. Wang, “In vivo full rang complex Fourier domain optical coherence tomography,” Appl. Phys. Lett. 90, 054103 (2007).
20. R. K. Wang, “Fourier domain optical coherence tomography achieves full range complex imaging in vivo by introducing a carrier frequency during scanning,” Phys. Med. Biol. 52, 5897–5907 (2007). [CrossRef] [PubMed]
22. R. K. Wang and S. Hurst “Mapping of cerebrovascular blood perfusion in mice with skin and cranium intact by Optical Micro-AngioGraphy at 1300nm wavelength,” Opt. Express 15, 11402–11412 (2007). [CrossRef] [PubMed]
24. American National Standards Institute, American National Standard for Safe Use of Lasers: ANSI Z136.1 (Laser Institute of America, Orlando, Florida,2000).
25. M. Wojtkowski, V. Srinivasan, T. Ko, J. Fujimoto, A. Kowalczyk, and J. Duker, “Ultrahigh-resolution, highspeed, Fourier domain optical coherence tomography and methods for dispersion compensation,” Opt. Express 12, 2404–2422 (2004). [CrossRef] [PubMed]
27. B. Baumann, M. Pircher, E. Götzinger, and C. K. Hitzenberger, “Full range complex spectral domain optical coherence tomography without additional phase shifters,” Opt. Express 15, 13375 (2007) [CrossRef] [PubMed]
28. R. A. Leitgeb, R. Michaely, T. Lasser, and S. C. Sekhar, “Complex ambiguity-free Fourier domain optical coherence tomography through transverse scanning,” Opt. Lett. 32, 3453 (2007). [CrossRef] [PubMed]
29. E. A. Bedrosian, “Product theorem for Hilbert transforms,” Proc. IEEE 51, 868–869 (1963). [CrossRef]
30. S. L. Hahn, “Hilbert Transformation” in The Transforms and Applications Handbook,A.D. Poularikas, ed., (CRC, 1996), pp 463.
31. V. X. D. Yang, M. L. Gordon, A. Mok, Y. Zhao, Z. Chen, R. S. C. Cobbold, B. C. Wilson, and I. A. Vitkin, “Improved phase-resolved optical Doppler tomography using the Kasai velocity estimator and histogram segmentation,” Opt. Commun. 208, 209–222 (2002). [CrossRef]
32. M. Mujat, R. C. Chan, B. Cense, B. H. Park, C. Joo, T. Akkin, T. C. Chen, and J. F. de Boer, “Retinal nerve fiber layer thickness map determined from optical coherence tomography images,” Opt. Express 13, 9480–9491 (2005). [CrossRef] [PubMed]
33. S. Jiao, R. Knighton, X. Huang, G. Gregori, and C. A. Puliafito, “Simultaneous acquisition of sectional and fundus ophthalmic images with spectral-domain optical coherence tomography,” Opt. Express 13, 444–452 (2005). [CrossRef] [PubMed]
34. R. K. Wang, “Directional blood flow imaging in volumetric optical micro-angiography achieved by digital frequency modulation,”Submitted to Opt. Lett. for publication.
35. M. Szkulmowski, A. Szkulmowska, T. Bajraszewski, A. Kowalczyk, and M. Wojtkowski, “Flow velocity estimation using joint Spectral and Time domain Optical Coherence Tomography,” Opt. Express 16, 6008–6025 (2008). [CrossRef] [PubMed]