Vessel skeleton-based vasodynamic optical flow for video deshaking in functional OCT angiography

Functional OCT angiography (fOCTA) enables non-invasive detection of retinal neurovascular coupling (rNVC) function. However, involuntary eye movements during dynamic imaging can displace the field of view, thereby challenging the accurate quantification of rNVC function at the single-capillary level. In this Letter, we develop a vessel skeleton-based vasodynamic optical flow (VDO flow) algorithm designed for fOCTA video stabilization and vasodynamic preservation. This method not only registers dynamic OCTA images with high speed (∼0.39 s per frame) and high accuracy (63.4% vascular skeleton recombination rate) but also effectively preserves critical vasodynamic responses. By applying VDO flow, we could precisely detect rNVC dysfunction at the single-capillary level in early diabetic retinopathy, even in the absence of significant changes in retinal vascular morphology. This advancement facilitates the clinical application of fOCTA.