Abstract
Purpose: :
To investigate perfused capillaries from adaptive optics scanning laser ophthalmoscope (AOSLO) videos using a technique for increasing flow contrast.
Methods: :
AOSLO videos were acquired using either 532 nm or 840 nm wavelengths. Capillary blood flow, including motion of individual leukocytes, was directly visible in the videos without the use of extrinsic contrast agents such as fluorescein. We developed and applied computer vision algorithms for blood flow identification and analysis. Videos were stabilized with intra-frame corrections for distortions due to eye motion. Flow contrast through perfused capillaries was increased by calculating the variance image of division videos. Specific capillary segments were selected for more detailed analysis. We selected segments that had leukocytes flowing through them and generated spatiotemporal intensity plots along the segment centerlines. By extracting traces from spatiotemporal plots, leukocytes could be labeled in the original video. Thus, each speed measurement could be individually verified.
Results: :
Vessel perfusion maps were calculated for videos of both wavelengths. No vessels were detected in the foveal avascular zone as imaged by overlapping videos in the fovea and parafoveal regions. We were able to identify additional perfused capillaries that were difficult to visualize in the original video. To investigate leukocyte behavior, vessel segments were selected, spatiotemporal plots were generated, and traces from individual leukocytes were extracted. Histograms of leukocyte speeds, derived from the slopes of each trace, were generated. To verify that each trace corresponded to a leukocyte, leukocyte positions were labeled in the original video.
Conclusions: :
Increasing flow contrast is an important first step in dynamic analysis. One needs a complete map of the vasculature in the analysis region for proper hemodynamic analysis. Otherwise, it is easy to miss non-obvious capillaries. Using spatiotemporal plots, leukocyte tracking can be applied along curved vessels at the level of the smallest capillaries. Future research includes the development of additional quantitative hemodynamic measures.
Keywords: imaging/image analysis: non-clinical • image processing • blood supply