Machine Learning Can Predict AMD Progression

Supervised machine learning models can help determine risk for AMD progression.

Supervised machine learning models can help predict progression of age-related macular degeneration (AMD) among patients at high risk and determine significant contributors, according to findings presented at the American Academy of Ophthalmology (AAO) 2022 annual meeting, held in Chicago from September 30 to October 3. The researchers identified several significant contributors to AMD progression among patients at high risk. These include: presenting and final visual acuity, any history of glaucoma, endophthalmitis, cataract, atrial fibrillation and use of biologicals (all P <.01). 

The study sought to quantitatively determine the significant contributing factors in AMD progression, as well as assess different machine learning models on predicting AMD progression in a high-risk patient population. 

To conduct this study, researchers collected the clinical data of 257 patients (169 with dry-to-wet AMD, 88 with nonprogressed AMD). To determine statistical feature importance, Chi-squared and ANOVA analysis were used. The investigators evaluated machine learning models with and without weights using the identified contributing factors to predict AMD progression using stratified 5-fold cross-validation. 


Park C, Bandhey, H, Amason J, Goa Q, Pajic M, Hadziahmetovic M. M.. Exploring contributing factors and supervised machine learning models for predicting progression of AMD. Poster presented at: The American Academy of Ophthalmology (AAO) 2022 annual meeting; September 30-October 3; Chicago. Abstract PO343