Authors (including presenting author) :
Au CL(1), Shum GTH(2), Chong SSY(1), Leow PL(1), Ko CKL(1)
Affiliation :
(1) Department of Ophthalmology, Tung Wah Eastern Hospital, Pamela Youde Nethersole Eastern Hospital (2) Department of Medicine (Endocrinology), Pamela Youde Nethersole Eastern Hospital
Introduction :
Artificial intelligence (AI) is popular in ophthalmology, and medical authority approved diabetic retinopathy (DR) screening systems are commercially available. Up to 2022, only one US FDA-cleared AI system for DR screening is available for use in Hong Kong, China.
Objectives :
This study aims to evaluate the real-world performance of the approved AI system in Hospital Authority clinical service population in the Hong Kong East Cluster.
Methodology :
Fundus photos taken in PYNEH medicine endocrine team DR screening program in 2020-2021 were studied. Primary outcome is to calculate the sensitivity and specificity (Se/Sp) of referable DR; and secondary outcome is to calculate the Se/Sp of detecting any DR or diabetic macular edema (DME). Ethics approval reference number was HKEC-REC-2022-019.
Result & Outcome :
The Se/Sp for referable DR was 96.2 & 94.7 respectively. The Se/Sp for detecting any DR or DME was 92.7 & 93.7 respectively. The studied system is a reliable AI tool for DR screening service in Hospital Authority settings. Usage in clinical service should be considered.