Segmenting and Labeling blood vessels in choroidal Haller's layer: A multiple target tracking approach

In ophthalmology, monitoring of choroid health assumes significance as various diseases, including age-related macular degeneration, tend to affect choroidal vasculature early. However, associated changes are often minute, and it remains a challenge to locate those. The traditional method, where clinicians glance through multiple 2D OCT images to make a diagnosis, is often imprecise and unreliable. Hence, it is imperative to develop technology-assisted reliable methods to detect and precisely locate minute structural changes in 3D. The present paper takes an initial step towards meeting such imperative. Specifically, we adopt a multiple target tracking approach to trace blood vessel systems of relatively large diameters belonging to Haller's layer. We obtained high accuracy in terms of Dice coefficient for synthetic images, and subsequently presented our results for clinical OCT images. Our method facilitates localization of anomalies at the scale of blood vessels, potentially revolutionizing future clinical practice.