AEYE-DS

The AEYE-DS is an artificial intelligence-based software designed for healthcare providers to automatically analyze retinal images from compatible fundus cameras (Topcon NW400, Optomed Aurora) for the detection of more than mild diabetic retinopathy in adult diabetic patients. Clinical studies have demonstrated its diagnostic performance, showing high sensitivity, specificity, and imageability, supporting its role as a screening tool for this condition.

The AEYE-DS is a retinal diagnostic software device utilizing artificial intelligence to automatically detect more than mild diabetic retinopathy (mtmDR) in adults diagnosed with diabetes who have not previously been diagnosed with diabetic retinopathy. It is intended for use by healthcare providers and functions by analyzing fundus retinal images acquired using compatible cameras, specifically the Topcon NW400 and the Optomed Aurora. The system architecture comprises five software components: Client (local user interface), Service, Analytics, Reporting and Archiving, and System Security, with the latter components typically server-based and accessed via a secure internet connection. The Client module guides image acquisition and transfer; the Service module manages communication and image transfer to the Analytics module, which processes images and returns diagnostic information regarding image quality and the presence or absence of mtmDR. Clinical performance for the AEYE-DS with the Optomed Aurora camera was demonstrated in two prospective, multi-center studies. In Study 1 (317 subjects), fundus-based sensitivity was 92% and specificity was 94%, with a Positive Predictive Value (PPV) of 68% and Negative Predictive Value (NPV) of 99%. In Study 2 (362 subjects), fundus-based sensitivity was 93% and specificity was 89%, with a PPV of 53% and NPV of 99%. Imageability was 99% in both studies. The device successfully identified all participants with ETDRS level 43 or higher in both studies. A precision study demonstrated high intra-operator repeatability (99%), between-operator reproducibility (95%), and between-device reproducibility (97%) with the Optomed Aurora camera. A human factors validation study confirmed ease of use and user manual comprehension by potential users after initial training.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the AI algorithm differentiate between 'more than mild diabetic retinopathy' and milder forms or other retinal pathologies visible in fundus images?; What are the specific image quality parameters the AEYE-DS software uses to determine if an image is 'insufficient quality,' and how might this impact the workflow and rescan rates in a busy clinical setting?; Given the variation in Positive Predictive Value (PPV) between Study 1 (68%) and Study 2 (53%) using the Optomed Aurora camera, what are the potential contributing factors to this difference, considering the study populations were not enriched?; The document mentions that 'novice operators' obtained fundoscopy images. What level of training was provided to these operators, and how does this compare to the expected training for typical healthcare providers intended to use the AEYE-DS system?; Study 2 indicated a significant effect of age on performance, with the lower age group (<55 years) showing better results. What are the hypothesized reasons for this age-related performance difference, and are there specific considerations or adjustments recommended when using the device with older patient populations?