Project Title 

South Western Eye and Diabetes Deep Learning Algorithm (SWEDDLA) study

Chief Investigator:

Dr Shweta Kaushik


Professor David Simmons, Dr Chee L. Khoo, Dr Marko Andric, Dr Kate McBride, Dr Jason R. Daley, Ms Xingdi Wang, Dr Vallimayil Velayutham, Dr Uchechukwu Levi Osuagwu

Aim of the project

To test whether an Artificial Intelligence algorithm, created in Australia, can diagnose vision-threatening diabetic eye disease with similar accuracy to eye specialists.


We took retinal colour photographs and retinal scans of persons with diabetes, and compared the grading of photographs by the Artificial Intelligence algorithm we created with the grading by eye specialists.

Key Results

To date we have compared images from 536 people with diabetes, including 35 children. We found 28.7% of the study population had any degree of diabetic eye disease and 5.8% had vision threatening diabetic eye disease. The algorithm was able to diagnose diabetic retinopathy with 91.8% accuracy from retinal photographs and 97.5% from retinal scans. 94.4% of participants would undergo retinal imaging again.


Our algorithm had excellent accuracy for diagnosing diabetic eye disease amongst the cohort that we have recruited so far.

Clinical Implications

This algorithm may be clinically deployable after further point-of-care testing. Our algorithm is unique as it includes children and combines information from retinal photographs and scans.


Daley JR, Wang X, Simmons D, Osuagwu UL, Vellayutham V, Khoo CL, Heydon P, Liew G, Andric M, Kaushik S. Development of a deep learning algorithm for provision of a South Western Sydney diabetes retinal screening service. [paper] The Royal Australian and New Zealand College of Ophthalmologists (RANZCO) 52rd Annual Scientific Congress, Brisbane, February 2022.

Wang X, Daley JR, Simmons D, Osuagwu UL, Vellayutham V, Khoo CL, Heydon P, Liew G, McBride K, Andric M, Kaushik S. Establishing a diabetes retinal screening service in South Western Sydney: Patient satisfaction with retinal imaging and the correlation between diabetic retinopathy and quality of life. [poster] RANZCO 52rd Annual Scientific Congress, Brisbane, February 2022.

Lay summary of outcomes

Diabetic eye disease is highly prevalent in South Western Sydney, greater than reported in other areas of Australia. Our Study’s preliminary findings suggest that an Artificial Intelligence program can bring specialist-level accuracy to diabetes hospital and general practitioner clinics, meaning patients can have their diabetes management optimised at the same time as their eye test.