Embedded Files
Stabilization
In order to elucidate the importance of physical stability, we tested recording with and without the eyepiece bracing the device to the subject’s eye. The instability under each condition was understood using the average frame-to-frame movement in the x and y direction as a proxy measurement. To put it another way, this looks at the relative movement between successive frames in a sort of mock-derivative of the position graphs — instances where the frame movement spikes imply little to no overlap between frames
Figure 1: Graph depicting the vertical and horizontal stabilization of the recordings with and without the eye piece. You can see that stabilization is significantly improved when recordings were taken with the eyepiece.
By tracking the x and y position change for each frame in a recording relative to the first frame, we can visualize the effect of digital stabilization. This is depicted in Figure 2, which shows the correction of small eye movements performed by the stabilization software.
Figure 2: Graph depicting the position correction done by the stabilization algorithm. Frame position relative to the first frame is shown for both horizontal (red lines) and vertical (blue lines) directions. The dotted lines indicate the position of the raw video recording which is adjusted to the solid line to produce the stabilized frames.
Micro-vessel Measurements
In terms of microcirculation analysis, vessel diameter and axial velocity within the vessels are the most important measurements as they are used for calculating blood volume (Q) and wall shear rate (WSR) results, which can play a role in clinic diagnosis and risk assessment of CVD.
1. Axial Velocity Calculation
For axial velocity calculation, a heat map of velocities on the captured image will be generated that estimates the axial velocities distributed along the blood vessels. The approach for the axial velocity calculation is a pairwise comparison of light intensity levels of each pixel with its neighbors in the image sequence, which are shifted forwards or backward in time in one-by-one frames. Then, the time-shifted pixels clumped together which produce the greatest similarities to the reference clumped pixels from the previous frames provide a distance the particular clump in the blood vessel has traveled during the inter-frame period.
Figure 3: Velocity heat map for estimating flow speed of particles flowing through the microvasculature. Each colored point shows the velocity of an object in a blood vessel based on pixel changes across consecutive frames.
2. Diameter Calculation
Blood vessel diameter is estimated using the EDT function available in MatLab. The Euclidean distance from the targeted pixel to its nearest non-zero pixel, within the binary image of specific vessels, will be represented for the pixel of EDT. As a result, the central line of the vessel will be indicated as the central EDT values, pointing out the value of the vessel radius along the longitudinal axis. Then, the average of diameters along a certain amount of the vessel length would give the mean vessel diameter27.
Figure 4: Simulation for vessel diameter estimation: (a) three vessels are simulated with diameter values of 25.3 pixels, 16.5 pixels, and 8.3 pixels, respectively; (b) the vessel centerlines are overlaid on the binary image; (c) the EDT of the binary vessel image. The central EDT values are approximately the same as these above values, which are 25.9 pixels, 16.6 pixels, and 8.6 pixels, respectively. DOI: 10.1016/j.mvr.2019.103907
Page updated
Report abuse