Integrated learning methods have been developed so far, and there are still great challenges in automatic cataract detection and computation tasks. In order to solve the problem of low accuracy and sensitivity of cataract grading diagnosis, this paper proposes an improved integrated learning method, EasyEnsemble, for IOL calculation and cataract diagnosis by improving the Adaboost algorithm with selected generations on the basis of undersampling. The cataract ultrasound dataset is selected and compared and analyzed with other methods, and the results show that the AUC, ACC, TPR, and TNR of this paper’s method are around 0.9, and its accuracy and sensitivity are much higher than that of other existing methods. And the experimental results based on the eye ultrasound image dataset show that the method integrated in this paper can adaptively focus on the abnormal region in the eye where cataract lesions occur, with better feature selection, and can more accurately characterize the cortical cataract.
