Use of Artificial Intelligence for Mass Screening of Diabetic Retinopathy in civilian Ukrainians during the full-scaled war

Abstract

PURPOSE

To create a patient-centric environment for early detection of DR with AI-driven solutions and increase the percent accuracy over 90% for mass screening of diabetic retinopathy in civilian Ukrainians during the full-scaled war.

SETTING / VENUE

SI"The Filatov Institute of Eye Diseases ans Tissue Therapy of the NAMS of Ukraine", Odesa, Ukraine Check Eye LLC, Kyiv, Ukraine Ukrainian Diabetic Federation, Kyiv, Ukraine

METHODS

The first stage – learning of the neural network. MedTech startup CheckEye has partnered with the Filatov Institute to conducted training of our proprietary neural network. There were used 12,000 color fundus images to train it to determine the stages and severity of DR. The second stage is the screening of patients including internally displaced people from Eastern and Northen regions of Ukraine with diabetes mellitus using the cloud storage with artificial intelligence platform in Chernivtsi region (Western part of Ukraine). People visited screening point in outpatient department or mobile screening point (equipped bus) to take a fundus photograph. Photos were taken by trained non-medical or junior staff using a standard non-mydriatic fundus camera. The platform analyzed photographs of the patient's eye fundus, making diagnosis of DR in few seconds. Quality control was performed by an expert group of ophthalmologists from the The Filatov Institute.

RESULTS

341 patients (682 eyes) with diabetes including internally displaced people were examined in October-December 2022. 111 patients (157 eyes) were determined the presence of DR for the first time. We achieved 92% accuracy in detecting DR and 82% accuracy in detection DR’s stage with AI-driven solution during the screening of a fundus which is better than average for humans doctors due to our comparisons.

CONCLUSIONS

There was created software environment for early detection of DR according to using solutions based on artificial intelligence. Our patient-centric environment is useful for mass screening of diabetic retinopathy in civilian Ukrainians including internally displaced people during the full-scaled war. Also it may cut shortage of healthcare by use of mobile screening point and acceleration of direction of patient to retina specialist if really needed.