Google image search offers rapid exposure to retinal diseases but has drawbacks
Google image search is a tool with acceptable accuracy for ophthalmologists who seek rapid exposure to retinal diseases, but it has significant limitations as a primary learning source, according to a speaker.
At the virtual Wills Eye Conference, Lucy Cobbs, MD, presented a study that she and colleagues conducted to assess the accuracy and quality of Google image search results for retinal diseases.
A Google search for 15 common and 15 rare retinal diseases found that the majority of images were fundus photos, making up 83% of the common retinal disease results and 81% of the rare retinal disease results. Less than 20% of images were annotated to identify pathological features.
Photo accuracy was graded based on the opinions of two expert graders and the context of the source website.
Of 300 images, 4% were incorrect (12 images, six common diseases and six rare diseases). Seven images showed the wrong disease, four images showed a normal fundus, and one image showed a fellow eye not affected by disease. Whereas 24% of all images were from peer-reviewed sources, 28% of images of common retinal diseases were from private or commercial practice websites, and 31% were aimed at patient education.
Only 17% of all images were annotated to show more in-depth features, Cobbs said.
“Particularly for common retinal diseases, non-clinical images were frequently included in the top results,” Cobbs said. “Only 24% of images were from peer-reviewed sources, and many images came from private or commercial practice websites that tended to target a patient audience.”
Collapse
Cobbs L, et al. Google image search as a tool to learn retinal diseases. Presented at: Wills Eye Conference; March 12, 2021 (virtual meeting).
Read more about