The invention provides an eye fundus image lesion screening system, and the system comprises a storage module deployed in a cloud server; an acquisition module for acquiring two or more eye fundus images from different eyes of the same person and uploading the eye fundus images to the storage module; and an inputting module which is configured to input the information f a tested person corresponding to the eye fundus images and associate the information with the eye fundus image, and upload the information of the tested person to the storage module; and a screening module which is used for extracting the eye fundus images and the information of the tested person from the storage module, and automatically interpreting whether the associated fundus images have the lesions or not through a deep-learning method based on the convolutional neural network so as to detect the presence of the lesions to generate a screening result. According to the present invention, the system can facilitate the popularization and promotion of fundus image lesion screening, and perform the interpretation of the eye fundus images by the deep learning method to generate the screening result, thereby reducingthe burden on an operator and improving the screening efficiency.