HK PolyU and OPTICAL 88 Release Initial Results of AI Diabetic Retinopathy Screening

In an effort to alleviate the burden on public primary eyecare departments, The Hong Kong Polytechnic University (PolyU) and OPTICAL 88 have pioneered a diabetic retinopathy (DR) screening project. This project is part of the “Smart primary healthcare and eyecare service: From data to algorithms and real-world solutions,” supported by Prof. Mingguang HE, PolyU Chair Professor of Experimental Ophthalmology at the School of Optometry. Prof. He, also a Global STEM Scholar under the Global STEM Professorship Scheme and Henry G. Leong Professor in Elderly Vision Health, has developed a low-cost, portable, self-testing retinal fundus camera integrated with a bespoke artificial intelligence (AI) system. This innovation has been tested on 320 eligible individuals, demonstrating an impressive accuracy rate of nearly 90% in detecting referable DR. Of these, 13 positive cases were identified, with 5 confirmed as diabetic retinopathy after further evaluation by ophthalmologists. The team has conducted telephone interviews and surveys to assess the diagnostic accuracy, efficiency, and practicality of the system, aiming for broader implementation across Hong Kong.

Fundus Screening of 50 to 59-year-old Saw Success Rates of Up to 97%

The research team conducted DR screening for eligible personnel across OPTICAL 88 outlets at Cityplaza, Telford Plaza, and The Optometry Clinic at The Hong Kong Polytechnic University. During the examination, a shot of either eye clearly showed the participant’s fundus, which were graded for risks of diabetes. Success was classified as det

ection within 3 shots. The study saw that the screening resulted in nearly a 90% success rate; 100% for participants under 50 years old; 96% for those aged 50 to 59 years old; and 69% for participants who are 70 years old or above. Evidently, the success rate declines with age. 11% of participants saw unsuccessful fundus photographs; while more than half of the participants suffered from small pupils or cataracts. Nearly 30% of participants were unable to obtain gradable fundus photographs within 3 shots; and about 10% of participants were unable to take the photographs due to their inability to focus on the target within the fundus camera. 13 positive cases were identified during the screening: symptoms were found in their fundus photographs. 5 participants were diagnosed with DR after further examination by the research team at the School of Optometry at The Hong Kong Polytechnic University.

AI Fundus Screening: Simple & Convenient – Enhance Optometrists’ Efficiency and Public Awareness of Optical Health

Ho Kai Wing, OPTICAL 88 Registered Optometrist (Part I) involved in the project, says “The performance of the AI DR screening was exceptional. Compared to traditional DR screening, the AI version is simpler, faster and accurate – thus more appealing to those who may be less aware on ocular health. Pupil-diluting eye drops aren’t necessary for the screening, and the process can be done through the AI system at the optical shop to ensure early detection.

He adds, “Optometrists recommend everyone should undergo a comprehensive eye examination once a year to ensure early detection on eye diseases. However, awareness for ocular health is still low in Hong Kong. The latest AI DR screening system hopes to not only increase the quality and efficiency of the work of optometrists, but it also educates the importance of regular comprehensive eye examinations.”

AI Portable DR Camera Improves Primary Eyecare in Hong Kong
Expanding the Network of Small and Low-Cost Self-Service Optical Screening

In addition to convenience for patients and eyecare professionals, the AI DR screening system is also cost-effective. Each screening takes only 5 minutes, a substantial reduction from the 30-minute traditional DR exams. The AI DR screening can process 2000 patients every month, while the latter can only process approximately 400. The former only costs 46% of the traditional method. To process 1000 patients, for example, the traditional exam requires approximately 2 months; while the AI DR screening needs only 10 days, meaning the screening of 1.5 million Hong Kongers aged 65 or above would only take 3 years, rather than 20 years in the traditional method.

Prof. Mingguang HE, PolyU Chair Professor of Experimental Ophthalmology of the School of Optometry, says his DR screening method combines technology and AI to solve 3 major issues in optical screening, thus realising the possibility of conducting optical screening in general outpatient clinics and optical shops. First, the small and low-cost cameras have reduced the cost issues for clinics. Second, the self-service screening process reduces dependency on technicians. Third, the AI diagnostics technology aids technicians in producing timelier and more accurate reports. Looking forward, AI diagnostic technology will not only be able to assist in the diagnosis of diabetic retinopathy, but also help diagnose a wider range of eye diseases, and offer specific treatments, thereby making these screenings more accessible, affordable and accountable.

Mr Joshua S. Kanjanapas WONG, Executive Director & Co-CEO of OPTICAL 88 says, “Helping our customers to maintain good eye health  is the top priority at OPTICAL 88. We have joined hands with The Hong Kong Polytechnic University to launch the AI DR screening research project with the aim to not only enhance the quality of our professional eye examinations, services and products, but to also promote primary  eye care in Hong Kong. The project can screen out patients with DR signs in a timely manner, so they can seek for medical consultation and treatments when necessary. This project had also collected useful data that can be used to evaluate the benefits of AI technology in primary eyecare. It can also promote the usage of self-service eye health screening tools, and provide insights for the future development of technology in eye care.”