Abstract
Purpose: :
To construct high contrast image of capillaries and to robustly calculate blood cell velocity from adaptive optics scanning laser ophthalmoscope (AOSLO) video with blink or involuntary eye movements using abnormal frame information automatically detected in image warping.
Methods: :
Our improved blood flow analysis method consists of image warping, capillary segmentation, and blood cell velocity calculation. In image warping, we automatically detect abnormal frames with low intensity, motion artifacts, or large translations, based on statistics of intensity, similarity measures, and translation values. For capillary segmentation, we calculate variance throughout frames except abnormal frames for each pixel position. To calculate blood cell velocity, we produce spatiotemporal image across the capillary region followed by trajectory detection of blood cells. If a detected trajectory is located next to abnormal frames, we weight the likelihood value for the trajectory to improve detectability. In this study, we recorded AOSLO videos using Canon prototype system with a high wavefront correction efficiency using dual liquid crystal phase modulator. The imaging light wavelength was 840 nm and the frame rate was 64 Hz. The scan area at the parafovea was 1.4 × 2.8° and was sampled at 200 x 400 pixels. The videos were recorded for 2 seconds. The proposed segmentation method was applied to 20 pairs of videos (with / without abnormal frames) for 4 healthy subjects. We investigated coincidence rate between the segmented capillary region from video with abnormal frames and that from video without them. The proposed trajectory detection method was applied to 10 simulated spatiotemporal images.
Results: :
The coincidence rate on segmented capillary area was 74.4% on average, which is better than in case of all frames including abnormal frames are used for capillary segmentation (48.9%) because of the automatic abnormal frame exclusion technique. In velocity measurement, as detectability of a short trajectory located next to abnormal frames was improved, the trajectory detection number increased by 27.8% on average compared with the result by conventional line detection method.
Conclusions: :
Our blood flow analysis method improved contrast of retinal capillary image and detectability for the short trajectory of blood cells from AOSLO video with blink or involuntary eye movements using automatically detected abnormal frame information.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • retina • image processing