Share on Facebook
Share on Twitter
Share on LinkedIn
Share by Email
Print
|
This article is part of Ophthalmology Advisor’s conference coverage from the 2021 meeting of the American Academy of Ophthalmology, held in New Orleans from November 12 to 15, 2021. The team at Ophthalmology Advisor will be reporting on a variety of the research presented by the ophthalmology experts at the AAO. Check back for more from the AAO 2021 Meeting. |
Presenters at the American Academy of Ophthalmology 2021 meeting reported that, using a machine learning approach, they were able to successfully predict diagnoses of vitreoretinal lymphoma (VRL), acute retinal necrosis (ARN) and endophthalmitis. The study employed a random forest algorithm that was developed using 28 immune mediators in the aqueous humor. The results show the predictive power of machine learning in ophthalmology.
The investigators applied the machine learning algorithms to aqueous humor immune mediator levels and combined with clinical data to predict the diagnosis of 17 intraocular diseases. The study included 512 eyes. The algorithms were evaluated by stratified κ-fold cross-validation where 28 aqueous humor immune mediators were measured with cytometric bead array and 7 types of clinical data were subject to machine learning.
Of 5 machine learning models, random forest was deemed the most successful in classification accuracy.
Top key factors were interleukin-10 for prediction of vitreoretinal lymphoma, and interleukin-6 for endophthalmitis.
|
Visit Ophthalmology Advisor’s conference section for complete coverage of AAO 2021 meeting. |
Reference
Usui Y, Nezu N, Kinya T, et al. Machine learning for intraocular disease prediction based on aqueous humor immune mediator profiles with clinical data. Paper presented at: American Academy of Ophthalmology 2021 Annual Meeting; November 12-15, 2021; New Orleans. Abstract PA012.
