Tuesday, April 29, 2025
4/29/2025
Smartphone-based Retinal Imaging for Diabetic Retinopathy Detection in Egypt - Project Summary: Diabetes affects 537 million adults globally in 2021, with 75% residing in low- and middle-income countries (LMICs). Diabetic retinopathy (DR) is a common complication of diabetes and manifests as retinal vein bleeding. Almost 10% of people with diabetes can progress to vision threatening DR without any symptoms so it is crucial to be evaluated by retinal screening and referred to a doctor. Since bulky and expensive fundus cameras are not suitable for fast DR screening in LMICs, there is an urgent need for portable and inexpensive retinal imaging systems. Recent technical innovations make use of smartphones as imaging devices to capture retina images but existing automated algorithms for image analysis are only designed for high-quality fundus cameras. Therefore, we propose an innovative provider-performed solution for smartphone-based retinal imaging to analyze images using deep learning for DR screening in an LMIC, as a case study in Egypt (DR prevalence of 20%). The primary objective of our study is to contribute to efforts to improve the technical capability and clinical practice for DR screening in Egypt, increase the rate of access to DR imaging, ultimately reduce the blindness. We intend to achieve this objective by pursuing two specific aims in the R21 phase to establish initial feasibility. First, we will attach a retinal imaging device to smartphones, creating a portable and cost-effective imaging system while optimizing image quality through contrast and sharpness enhancements. Second, since multiple images will be captured from smartphone-based retinal imaging systems to cover whole retina due to limited field of view, we will develop CNN-based deep learning algorithms to analyze the DR formation in these multiple images and fuse the decisions in score and feature levels. Upon successful completion of R21 milestones, the R33 phase will focus on further validation and effectiveness studies, with two specific aims. We will first utilize collaborative approaches and employ trustworthy WisdomNet architecture to improve model performance. Second, we will embed the knowledge into deep learning frameworks using attention map mechanism. Instead of using only labels to train the deep networks, we will also use the location of suspicious retinal regions during training to focus on these regions. This approach enhances model performance and provides explanatory insights crucial for assisting doctors in DR diagnosis and prioritizing cases for further examination. This project will enable fast DR screening in LMIC general healthcare facilities such as urban areas, remote clinics, and pharmacies. With this project, we will have the potential to distribute quality eye care to underserved areas lacking access. In the long term, we will expand this smartphone-based retinal imaging system to other LMIC and to screen for other eye diseases.
HHS’ Tracking Accountability in Government Grants System (TAGGS) website provides access to detailed descriptions of grants, loans, aggregated direct payments and other types of financial assistance awarded by the HHS Operating Divisions (OPDIVs) and the Office of the Secretary Staff Divisions (STAFFDIVs).