Referral for Disease-related Visual Impairment using Retinal Photograph-based Deep Learning: A Proof-of-Concept

The article titled "Referral for disease-related visual impairment using retinal photograph-based deep learning: a proof-of-concept, model development study" was recently published in Lancet Digital Health by IHPC and Singapore Eye Research Institute (SERI), in collaboration with other local and international partners. The current IHPC researchers involved in this work are Mr Gabriel Tjio, Dr Li Shaohua, Dr Xu Xinxing, Dr Rick Goh and Dr Liu Yong.

Visual impairment is a major public health problem associated with reduced quality of life and increased risk of frailty and mortality. Analysis of worldwide data in the two decades from 1990 to 2010 suggests that 60% of visual impairment is disease-related and cannot be corrected with spectacles; instead, disease-related visual impairment require medical attention and possibly even surgery in eye-care settings led by ophthalmologists. In Singapore, about 12% of those aged above 60 years (about 90,000 individuals) suffer from pathology-related visual impairment. Fortunately, if detected early, the potential disease-related vision impairment can be averted or slowed.

This article presents a proof-of-concept that a single-modality deep learning algorithm could be used to identify visual impairment due to different eye diseases, as well as the extent of the visual impairment; the information provided by the deep learning algorithm could then form the basis of referrals to ophthalmologists for further investigation and intervention. When validated using retinal images from the three main ethnic groups in Asia – Malay, Indian, and Chinese, the deep learning algorithm was able to identify indicators of vision impairment associated with different eye diseases. The long-term research vision is to develop and deploy a community screening system to provide efficient, automated, and pinpointed referral of patients from the community to tertiary eye hospitals, so that mass community screening could be carried out while simultaneously minimising the risk of over-burdening tertiary hospitals with unnecessary referrals.  To the best of our knowledge, this is the first study to show the use of only a single macular-centred retinal photograph for identification and referral of eyes with disease-related visual impairment.

This work is a core component of IHPC's strategy in Digital Ophthalmology to deploy AI solutions in community care, primary care and tertiary care environments. The IHPC-SERI team intends to refine the deep learning algorithm, as well as to carry out the necessary adaptions for and testing under the potential deployment operational settings, so as to achieve the vision of deploying deep learning systems in community care facilities to better support the eye care needs of Singapore's aging population.

Read the full article published on Lancet Digital Health.