Using our method, we compared flow patterns between inlet and outlet of the OFT. The inlet position considered (Y coordinate) corresponds to the blue line in and ~0.6 mm in . At this position, the Doppler angle is about 73.8°, DF is 3.59, and the lumen cross-sectional area is approximately 0.34mm2, when the OFT walls are expanded. From our 4D scan protocol, at this position we have a sequence of 200 frames (2D images) of raw data. For our analysis, we selected a phase (time) in which the OFT had maximum flow velocity. This selected phase is shown with a white line in
–an M-mode image extracted (post-processed) from the acquired sequence of cross sectional images, along the middle line of the OFT lumen. For the phase selected, we had both a structural image  and a Doppler SDOCT velocity distribution image, from which the projection of the blood flow velocity in the axial direction (Vz) can be calculated. We corrected the velocity distribution within the OFT, Vz, using the calculated DF to get the absolute blood flow velocity distribution in the OFT . Further, blood flow velocity was quantified along a horizontal line  that approximately cut the OFT cross-sectional image in half [white line in ].
Fig. 6 Maximum blood flow velocity near the OFT inlet. (a) M-mode image extracted from the sequence of cross-sectional images (one row over time). White line in (a) denotes the phase (time) of the cross-sectional images (b) and (c); (b) OCT structural image; (more ...)
For the OFT outlet (corresponding position is shown in (yellow line) and ~0.25 mm in ), the Doppler angle is 68.9°, DF is 2.78 and lumen cross-sectional area is approximately 0.25mm2. We performed the same data processing procedures described for the OFT inlet position to get the absolute blood flow velocity and velocity profile in the OFT outlet position when the OFT was expanded and blood flow velocity was maximal (see
Fig. 7 Maximum blood flow velocity near the OFT outlet. (a) M-mode image extracted from the sequence of cross-sectional images (one row over time). White line in (a) denotes the phase (time) of the cross-sectional images (b) and (c); (b) OCT structural image; (more ...)
Comparing and , we observe that the blood flow direction is negative (blue) at the OFT inlet while the flow direction is positive (red) at the OFT outlet. This is because the OFT bends, and thus blood flows ‘up’ (negative direction) at the inlet and ‘down’ (positive direction) at the outlet. The maximum absolute blood flow velocity at the inlet is ~35 mm/sec  while at the outlet it is ~50 mm/sec . Thus the outlet of the OFT has a higher maximum blood flow velocity than the inlet, trend that was also observed in our ultrasound data (not published). This is probably because the OFT tapers toward the outlet and as a result the cross-sectional area of the lumen at the inlet, ~0.34mm2  is larger than that at the outlet , ~0.25mm2.
We also analyzed an image at the OFT outlet that showed maximum backflow velocity (when blood flows from the arterial system backs to the OFT). The position of this image (Y direction) is the same as that shown in and yellow line, but the phase is different [see
]. Comparing white lines in and , we observe that maximum backflow appears a little earlier than maximum forward flow. Maximum backflow velocity is ~18 mm/sec , which is smaller than maximum forward flow velocity. From our observations, backflow is generally larger at the outlet than at the inlet of the OFT. In some embryos, further, there is no backflow at the OFT inlet, such as in the case analyzed here.
Fig. 8 Maximum backward blood flow velocity near the OFT outlet. (a) M-mode image extracted from the sequence of cross-sectional images (one row over time). White line in (a) denotes the phase (time) of the cross-sectional images (b) and (c); (b) OCT structural (more ...)
Our method to calculate blood flow velocity, needs to segment the OFT boundary from SDOCT structural images to calculate the flow direction. Thus, the quality of the structural images is important for the accuracy of our method. At early stages of chicken embryo development, embryonic heart tissue is transparent. Furthermore, in order to make deep heart structures more clear we adjusted the position of the probe beam focus point to 800µm beneath the match point. This allows OCT to achieve relatively better defined structures at larger imaging depths. Image wash-out, however, is a phenomenon that deteriorates image quality. It appears as a shadow under the flowing medium in structural images affecting mainly the deeper heart OFT image boundaries. Image wash-out is determined by the flow velocity, especially the projection velocity in the Z direction (Vz), as well as the OCT imaging speed. In the current study, imaging speed was 47,000 A-scan per second, which determined the maximum measurable velocity was 12 mm/s (in probe beam direction). The early stage chicken embryo heart (HH18) is tiny and blood flow speed is not fast. Additionally, the Doppler angle of the blood flow in the OFT is ~70°, which makes Vz even smaller. Thus, the effect of image wash-out is not significant in our application and we can discern the deeper boundary of the OFT in most cases.
Doppler based flow measurement techniques (laser Doppler, ultrasound and Doppler OCT) always face the problems of Doppler angle correction. Generally, one representative longitudinal 2D image is used to set the Doppler angle in ultrasound. Similar methods were used in Doppler OCT [28
]. However, precision is the main issue of these methods, especially when trying to determine the Doppler angle of the beating embryonic heart. The chicken embryonic OFT at HH18 stage of development is a tube, but the shape of it is not regular.
and show different views of a segmented OFT of a chicken embryo. The OFT shows two bends in different directions, and the presence of these bends makes it difficult to get a good longitudinal image to determine the Doppler angle. For example,
shows a 2D longitudinal scan image, in which it is observed that the boundary at the left is cut (the OFT bends there). Using such longitudinal image to determine the Doppler angle, may introduce errors. Additionally, if the Doppler angle is measured ‘visually’ (e.g. on the screen), operator-related errors will be introduced. This is even worse for the case of the beating heart, because the Doppler angle is constantly changing due to the heart motion. Further, because DF
is sensitive to small variation in Doppler angle, especially when the angle is relatively big (~> 70°), a small imprecision in the Doppler angle measurement may affect the final result seriously. Comparing with the traditional method of measuring the Doppler angle, our method obtained Doppler angle by first reconstructing 3D images of the OFT at a specific phase of interest, then calculating the center line(thus including variations due to 3D bending), and finally calculating DF
from the obtained center line. As a result, our method of determining the Doppler angle and the associated DF
is more precise than traditional methods.
Chicken embryo OFT bending. (a) Side view of segmented OFT; (b) Top view of segmented OFT ; (c) Longitudinal 2D scan of OFT. The OFT is bending in two direction, so it is difficult to get a longitudinal cut of OFT.
In order to get the Doppler angle, Davis et al. [21
]acquired three-dimensional volume data of the chicken embryonic heart in vivo using SDOCT. Their 3D data were acquired over ~9.8 sec and processed by Amira software to render the heart surface, from which they measured the Doppler angle. An advantage of their method is the use of 3D data, which renders more precise values of the Doppler angle than using a 2D longitudinal image. The drawback of their method is that motion artifacts arise in the 3D image when the heart beats fast. For example, at HH18, the cardiac period of the chicken embryonic heart is about 0.3 to 0.5 seconds. Using our system, ~1.2 seconds are needed for a 3D scan (faster than the 9.8 seconds in the system of Davis et al). A 3D volume data set comprising multiple cross-sectional images along the OFT will then be imaged over 34 cardiac cycles, and therefore the volumetric reconstructions will not be accurate. Obviously, the measured Doppler angle from these kind of data sets are subjected to errors. In our method, to circumvent this problem and get accurate 3D volume data over the cardiac cycle (that is, 3D images that represent the geometry of the heart at different phases of the cardiac cycle) we use a 4D scan strategy and synchronization procedure [24
]. Therefore our method can be applied when the heart is beating.
Yeh et.al [29
] presented a method to measure three components of an arbitrary velocity vector based on SDOCT in 2007. A beam divider, which divides a probe beam into five independent view points and path length delays, was designed. By inserting the divider into the sampling arm and using certain phase-resolved algorithm, they quantified the three-dimensional velocity vector. The advantage of their method is obvious, i.e. it can achieve three-dimensional velocity by one scan. However, it is difficult to measure the chicken embryo OFT blood flow velocity using this method because of sample size requirements. In their method, one sample was shifted and imaged to five view points, i.e. five structural images were spread over the measurable range of SDOCT. In our SDOCT images, the deeper OFT borders were located at 1.5~2mm depth. That means that, even in ideal condition (neglect structures beneath the OFT), we need 7.5~10mm measurable range SDOCT system in order to successfully use the method described by Yeh et al. Our system can only provide at most 4.6mm measurable range, and thus structural overlap is unavoidable, and the method cannot be successfully implemented in our current system for the measurement of the heart OFT at HH18.
As we mentioned before (Section 3.4), Doppler SDOCT uses phase differences (
) between two adjacent A-scans to calculate the projected blood flow velocity Vz
]. However, the range of
is [-π,π]. Thus, initially, as Vz
increases proportionally, until
reaches the value of π and
suddenly jumps to -π. This is called phase “wrapping” and can lead to incorrect velocity measurements. From Eq. (4)
we can see that the measurable range of the projection velocity Vz
is limited by the camera scan speed (τ) since λ( = 1321.5nm) and n
( = 1.3 in tissue) are constants. For our system, the measurable range of Vz
is [-12,12] mm/s when the camera runs at maximum rate (47,000Hz). Thus, if the projection velocity exceeds this range, wrapping will appear.
shows an example of the phase wrapping in the chicken embryonic OFT at forward blood flow phase. The wrapping (blue) area corresponds to the region in which Vz
exceeds 12mm/s. According to the color bar, we can see that blue denotes negative velocity, and this is obviously incorrect. We introduce a constant phase (π or -π, depending on the flow direction) to the second A-line before the phase difference calculation, after phase difference calculation we subtract this constant phase back. Using this method, we can shift the phase difference (
) range to [0, 2π] (or to [-2π, 0]. Consequently, the measurable range of the projection velocity is doubled, i.e., [-24, 24] mm/sec. The unwrapping result is shown in , the wrapping area [blue ] has been corrected (red) and the corresponding Vz
are higher than 12mm/sec.
Phase unwrapping. (a) Forward blood flow with wrapping; (b) Forward blood flow with wrapping correct. Measurable range shifted from [-12, 12] mm/sec ((a))to [0, 24]mm/sec((b)).
Our developed method to measure blood flow velocities is suitable to study blood flow dynamics in the hearts of developing embryos, when the heart is tubular. Because the method proposed calculates the Doppler angle from 3D volume images of the heart at specific phases during the cardiac cycle, variations in the orientation of the heart due to cardiac motion are taken into account. This allows for an accurate estimation of the Doppler angle and the Doppler factor, DF, in the three-dimensional space and over the cardiac cycle. The method works best when blood flows in the direction of the heart center line, and therefore when flow velocities normal to the centerline are negligible. Velocities perpendicular to the tubular heart wall could occur due to secondary flows and wall motions (contraction, expansion and rigid body motion) during the cardiac cycle. While a careful analysis might be needed to determine whether flow velocities normal to the center line are negligible, this is also a problem that other Doppler velocity measurements face. In the case considered here, for the calculation of blood flow velocity in the OFT of an HH18 chicken embryo, the OFT is expanded and wall motion is minimal (see and ). Even during backflow, when the OFT walls are expanding (see ), the wall velocity – as estimated from the M-mode image – is about 2 mm/sec, negligible compared to the maximum backflow velocity of ~18 mm/s. Further, velocity profiles were calculated along the X-direction (, , and ), where the effects of expanding wall motions in Vz are minimized. Because Doppler OCT gives the spatial distribution of blood-flow velocities, our method could be employed in the calculation of wall shear stresses in the developing heart using signals from scattered red blood cells.