Comparison of Ensemble Learning Algorithms for Cataract Detection from Fundus Images

Cataract is a clouding or opacity of the eye's lens that can cause vision problems. It is widely accepted that early detection and treatment can reduce the suffering of cataract patients and prevent visual impairment from turning into blindness. This paper compares studies on the use of ensemble learning algorithms for cataract detection from fundus images. Two independent feature sets as texture-based and sketch-based are extracted from each fundus image. Three basic learning models as decision tree (DT), back propagation neural network (BPNN) and sequential minimal optimization (SMO) are built on each feature set. Then, the ensemble learning algorithms of majority voting and stacking method are investigated to combine the base learning models for cataract detection. A real-world data set including fundus image samples with no cataract, mild, moderate, and severe cataract is used for training and testing. Experimental results show that good performance results from the stacking method, with texture-based features giving accuracy of detection at 95.479%.