Skip to main content
main-content
Top

08-14-2019 | Retinopathy | News

Offline, smartphone-based imaging accurately detects diabetic retinopathy

medwireNews: An offline, smartphone-based, nonmydriatic retinal imaging system accurately detects referable diabetic retinopathy (RDR) when used by minimally trained operators in a point of care setting, show data from a pilot study carried out in India.

“In a developing country such as India, nearly 70% of the population resides in rural areas, and a ratio of only 1 ophthalmologist per 100000 people is available for the care of the entire population,” write Astha Jain (Aditya Jyot Foundation for Twinkling Little Eyes, Mumbai) and co-authors in JAMA Ophthalmology.

“This study shows how an offline AI [artificial intelligence] algorithm can help address this lack of specialist access through automatic, instant grading of the retinal images, highlighting a possible solution for implementation of large-scale models for screening for RDR,” they add.

When tested among 213 individuals with diabetes who visited various dispensaries in Mumbai, the Medios AI system (Remidio Innovative Solutions Pvt Ltd, Bengaluru, India) had 100% sensitivity and 88.4% specificity for the diagnosis of RDR – defined as any retinopathy more severe than mild diabetic retinopathy, with or without diabetic macular edema – compared with ophthalmologist grading of the same images.

The corresponding values for any DR were 85.2% and 92.0%.

Jain and colleagues explain that Medios AI is an offline smartphone-based, automated analysis system for retinal images that is paired with a nonmydriatic retinal camera.

The system, which was used by dispensary workers with no professional experience in the use of fundus cameras, first captures an anterior segment photograph followed by three fields of the fundus (the posterior pole, nasal, and temporal fields). A second DR assessment mechanism that relies on two convolutional neural networks then generates a diagnosis by detecting DR lesions.

The researchers note that the specificity may have been lower than expected because 67% of 12 cases of mild nonproliferative DR were diagnosed as RDR by the automated analysis.

“This is because the offline AI was purposely not trained on mild nonproliferative DR images, to ensure high specificity in no DR and RDR diagnoses,” they write.

In addition, “[r]etinal lesions other than DR, such as retinitis pigmentosa, drusen, and retinal pigment epithelium changes, were overdiagnosed by the AI as RDR. Although incorrectly labeled as RDR, these cases likely would warrant a referral to an ophthalmologist,” Jain et al remark.

The investigators add that validation of the findings in a larger cohort is now underway.

In an accompanying commentary, TY Alvin Liu, from Johns Hopkins University in Baltimore, Maryland, USA, says the pilot study represents a “major advancement” in the field of fully automated screening for DR in the primary care setting.

He comments: “The relative low cost and high portability of smartphones, coupled with the lack of need for a working internet connection, dramatically increases the scalability and potential effects of such a screening program.

“Second, by embedding the screening algorithm into smartphones, the authors have pioneered a way to potentially bring DR screening to patients at risk, instead of relying patients to come to points of care to be screened.”

Liu concludes: “This paradigm-shifting approach to DR screening could greatly benefit rural populations in both developing and developed countries where the access to care is limited.”

By Laura Cowen

medwireNews is an independent medical news service provided by Springer Healthcare. © 2019 Springer Healthcare part of the Springer Nature group

JAMA Ophthalmol 2019; doi:10.1001/jamaophthalmol.2019.2923
JAMA Ophthalmol 2019; doi:10.1001/jamaophthalmol.2019.2883

Related topics