Smartphone based fundus camera for the diagnosis of retinal diseases

Abstract Fundus photography is necessary to monitor the progression of retinal diseases like diabetic retinopathy, age-macular degeneration (AMD) and glaucoma. Traditionally, fundus photography is done by a specialized fundus camera which costs tens to hundreds of thousand dollars and weighs around fifteen kilograms. In this paper, we develop a smartphone based fundus camera for the early detection of the retinal diseases for the people in destitute areas who do not have access to the costly fundus camera and ophthalmologists. A major challenge in the smartphone based fundus photography system without involving an ophthalmologist is to ensure that the captured fundus image content is sufficient for detecting retinal diseases. To obtain a perfect fundus image, it is essential that the photo is shot exactly at the time when the camera is perfectly positioned and focused. To address this issue, our system captures a video and then selects a small set of fundus images from the video that have the required features for detecting retinal diseases. The manual selection of the best fundus images from the video is again a cumbersome process and requires an ophthalmologist. We propose an automation technique to identify the best fundus images from the video with low processing overhead. We evaluate the effectiveness and efficiency of our automation technique using in real settings. Similar to the original fundus camera, our system dilates the pupil before capturing the video. However, in this paper, we also investigate the challenges for smartphone based fundus photography without dilating the pupil of the eye and show the future research directions.