AI may estimate BCVA from fundus photographs in eyes with diabetic macular edema
Click Here to Manage Email Alerts
Key takeaways:
- AI used fundus photos to estimate best corrected visual acuity in patients with diabetic macular edema.
- Visual acuity estimates could be combined with OCT central subfield thickness measurements.
Artificial intelligence may be able to estimate the best corrected visual acuity of patients with diabetic macular edema from fundus photographs, according to a study.
“This investigation found that AI can estimate BCVA directly from color fundus photographs of patients with DME, often within 1 to 2 lines on an ETDRS chart,” the study authors wrote. “These findings support use of AI systems to estimate BCVA using this method for patients undergoing anti-VEGF treatment of DME without refraction or subjective VA measurements.”
Image: Adobe Stock
The authors used deidentified color fundus images taken after dilation to train AI systems to perform regression from image to BCVA. The analysis included 7,185 fundus images of study and fellow eyes from 459 participants.
Baseline BCVA score in the study eyes ranged from 73 to 24 letters. The mean absolute error (MAE) for the testing set, which included 641 fundus images, was 9.66, with 33% of the values within 0 to 5 letters and 28% within 6 to 10 letters.
The MAE was 8.84 letters for BCVA scores of 100 letters or fewer but more than 80 letters and 7.91 letters for BCVA scores of 80 letters or fewer but more than 55 letters.
Going forward, the AI estimate of BCVA could be combined with OCT central subfield thickness measurements to help determine follow-up or re-treatment intervals.
“These findings open up possibilities of using AI in the clinic or with home monitoring to obtain BCVA efficiently and economically, decreasing the time and expense related to the trained personnel involved in managing DME around the world,” the authors wrote.
Collapse
Read more about
Click Here to Manage Email Alerts