Project: Vessel Segmentation in Retinal Fundus Images
The leading causes of irreversible blindness in the United States are mainly age-related macular degeneration, cataract, and glaucoma. Early detection of these diseases is important for patients to optimize their treatment benefits. Among many imaging modalities used for retina like Optical Coherence Tomography (OCT), fundus photography is one of the most commonly adopted methods for ophthalmologists to detect and assess symptoms of eye diseases. The existing algorithms for automatic fundus image analysis can segment and measure the retinal vasculature, which is a major indicator of various diseases. However, the published work obtains insufficient information about vessel branching patterns and related measurements. Therefore, our project will focus on developing a novel machine-learning algorithm to locate the vessels, specify branching networks and connectivity, and compute measurements for targeted pathologies with high accuracy. The results of our algorithm will be compared with standard references marked by trained clinicians, which are provided in the majority of fundus image databases. We can possibly apply the same technique to 3D OCT images to reconstruct 3D vessel structures, if our method can function properly for 2D fundus images.