Eamon N. Dreisbach , 2025-05-05 20:00:00
Key takeaways:
- A multimodal AI model determined if eyes were healthy, glaucomatous or glaucoma suspected.
- The model combined image and parameter data to make its predictions.
SALT LAKE CITY — Multimodal AI models may be useful to predict glaucomatous status of eyes, according to a poster presentation at the Association for Research in Vision and Ophthalmology meeting.
Andy Ng and colleagues created an AI model that combined image and parameter data to classify eyes of 652 participants as glaucomatous, glaucoma suspected or healthy control. They defined glaucomatous eyes as those with glaucomatous disc and visual field and glaucoma suspected eyes as those with ocular hypertension or suspicious disc with additional risk factors including high IOP, diabetes or ethnicity.
“What we saw was that a lot of the patients who were marked as glaucoma suspect status eventually were diagnosed with glaucoma,” Ng told Healio.
Overall, 339 eyes were classified as healthy, 139 as glaucoma suspected and 174 as glaucomatous. The average area under the curve of the model for healthy, glaucoma suspected and glaucomatous eyes was 98%, 90% and 90%, respectively.
“On average, our model did the best at predicting healthy eyes, but it also did a pretty great job at predicting if eyes were glaucomatous or glaucoma suspect,” Ng said.
Sources/Disclosures
Collapse
Ng A, et al. Multimodal AI model prediction of glaucoma status. Presented at: Association for Research in Vision and Ophthalmology meeting; May 4-8, 2025; Salt Lake City.
Disclosures:
Ng reports no relevant financial disclosures. Please see the study for all other authors’ relevant financial disclosures.
