Approved Research
Classifying retinopathies from fundus images and optical coherence tomography data using deep-learning methods
Approved Research ID: 85789
Approval date: February 16th 2023
Lay summary
This study trying to establish a deep learning system for retinopathies recognition, including polyarteritis nodosa retinopathy and diabetic retinopathy.
Diabetic retinopathy was caused by damage to the blood vessels of the retina due to consistently high blood sugar. It might cause no symptoms or only mild vision problems initially, but it can lead to blindness in end stages. Therefore, early diagnosis and timely treatment were necessary. Besides, polyarteritis nodosa (PAN) is a medium vessel vasculitis that may affect the eyes and the central neural system. About 10% of patients suffered retinopathy due to the disease progression, which may damage their vision severely. What's more, PAN retinopathy and diabetic retinopathy were caused by different etiology, making them present different morphologies in the fundus.
Because of the above reasons, this study is going to develop deep-learning systems, which use algorithms and statistical models to analyze pictures and data, for detecting retinopathies, and the project duration would be expected as 36 months in total. The research not only investigates the variety of morphologies in the fundus in different retinal diseases but also a great aim for early detection of retinopathy and timely treatment. In addition, due to its characteristics of being cheap, rapid, and easily obtained, the systems could become a nationwide routine screening for early detection and timely treatment, which improved patients' outcomes with better preservation of vision.