In this SBIR project, we propose EyeScreenBot, an end-to-end automated retinal im- age capture and analysis system, comprising a self-driven, robotic fundus camera plat- form for automated image capture and a deep learning-based image analysis engine for generation of automated screening outcome. With the large, growing, and aging popula- tion and the increased prevalence of diabetes, a large number of people are at risk for vision loss due to several eye diseases including diabetic retinopathy (DR), age-related macular degeneration (AMD), and glaucoma. Although eye screening is effective in re- ducing vision loss, there are not enough clinical personnel and eye-care experts for pop- ulation-wide eye screening. Recent advances with automated image analysis are helping alleviate the situation, but they are still limited by the need for good quality images of the patients captured by trained technicians or expensive retinal cameras equipped for auto- mated capture. EyeScreenBot will be developed to provide a truly end-to-end screening solution that is cost-effective and suitable for deployment in primary care clinics or op- tometrist sites, addressing both automated capture and subsequent automated analysis, all without the need for trained technicians or eye experts at the point of care. When deployed and commercialized, this device will rapidly aid scaling of eye screening for the masses, thereby having an enormous impact in improving the quality and accessibility of eye care and helping reduce preventable vision loss.
EyeScreenBot, an end-to-end automated screening system with intelligent image capture and analysis, will truly enable eye screening at massive scale, which is necessary and urgent since the population at risk for preventable vision loss due to retinal diseases (such as diabetic retinopathy) is growing at a staggering rate. Triaging and identification of at-risk patients will allow for timely intervention to prevent, slow, or even reverse the disease progression and loss of vision.
