Publisher: Euretina Abstracts

Authors: Alberto Piatti, Carlo Bruno Giorda, Francesco Romeo

Diabetic Retinopathy Screening with Artificial Intelligence Preliminary experience in Italian Healthcare System

Purpose:

New technologies, such as non- mydriatic cameras and artificial intelligence (AI) software, allow to recognize elementary lesions, making a first diagnosis of retinopathy, reducing the diagnostic burden on eye care specialists and time costs for patients.

We present a preliminary report of the performance of artificial intelligence RetCAD ™ ( Thirona Retina BV Nijmegen, The Netherlands). The retinal images were obtained with a true –color, confocal, fully automated non-mydriatic fundus imaging system (DRSplus ® Centervue S.p.a Italy.

Setting / Venue

ASL TO 5 is a Local Health District of Turin Metropolitan Area with 310.315 residents. In 2019, ASL TO5 performed an integrated care pathway dedicated to diabetic retinopathy. The clinical pathway develops technological improvements, like retinopathy screening with artificial intelligence.

Methods

Diabetic patients were enrolled for a prospective observational study attending their annual visit. Participants were aged 18 years or older with diagnosis of diabetes mellitus type 1 or 2. All patients provided written informed consent . The retinal images were obtained with a true –color confocal, fully automated non-mydriatic fundus imaging system (DRSplus ® Centervue .Sp.a, Italy in a primary care setting (District of Moncalieri) without that participants underwent pupil dilatation. Two photographic image fields were taken of each eye, one centered on the optic disc and the other on the macula . Afterwards, the images were sent through a digital interface ( iCare ILLUME by iCare Finland Oy, Finland) , to the artificial intelligence. Instantly iCare ILLUME provides a report available on personal computer or even on smartphone. Data security and privacy met the highest standards.

Results

337 patients were included in a sample calculation of sensitivity, specificity, positive and negative predictive value.

RetCAD provided a retinopathy grading according to International Diabetic Retinopathy Severity Scale. When grading was more than mild retinopathy , patients were referable and sent for a complete exam to an ophthalmologist . Mean (SD) participants age was 67 (±11,7) years (range 38 – 92 years). Out of 337 individuals, 13 (3,8%) had type 1 diabetes, 324 type 2 (96,2%). A total of 142 participants were women (42,4%) and 195 were men (57,6%).RetCAD was able to detect all patients with more than mild retinopathy (sensitivity rate of 100%), positive predictive value , indicating the percentage of patients with retinopathy among those with a positive AI result, was 94,8% In addition RetCAD showed a very low false positive rate: 4 cases out of 77, with specificity rate of 98% (95% confidence interval ± 0,02 ) Negative predictive value , indicating the percentage of patients without retinopathy among those with a negative AI result, was 97,4 %.

Conclusions

New Knowledge and technologies such as digital non mydriatic cameras and artificial intelligence offer important support in diabetic retinopathy screening. Among the main automated image assessment systems, RetCAD throught the digital interface iCare ILLUME demonstrated an excellent sensitivity in detecting more than mild retinopathy. In addition, specificity was very accurate with a low false positive rate. This AI system can improve DR screening and monitoring in people with diabetes by non-eye care professionals. Nevertheless, future research is required to address several challenges of automated image detection algorithms : for example medico-legal implications or management of other ophthalmic disorders like age macular degeneration, detected in diabetic patients undergoing screening for retinopathy.