In-Clinic Hurdles for AI-Based Diabetic Retinopathy Detection

Source: Medscape

Researchers found mixed results when they tested an artificial intelligence (AI)‒based system for detecting diabetic retinopathy (DR) in a real-world primary care setting in Midland, Australia, according to a Medscape report.

In the study by Yogesan Kanagasingam, PhD, from the Australian e-Health Research Center and the Commonwealth Scientific and Industrial Research Organization in Perth, Western Australia, and colleagues, the AI system was used to evaluate 193 patients and 386 images between December 1, 2016, and May 31, 2017. Images were then read by an ophthalmologist for comparison.

The AI system identified 17 patients as having clinically significant disease, which required referral to a specialist. Of those, two patients had true disease, and 15 were misclassified as having disease (ie, received false positive results). In addition, the AI system correctly identified all eight cases of mild DR.

Read the full article.


Related Content