DIABETIC RETINA IMAGE CLASSIFICATION METHOD AND SYSTEM BASED ON DEEP LEARNING

The present invention discloses a method and system for classifying diabetic retina images based on deep learning in the field of artificial intelligence technology. The method comprises the following steps: obtaining a fundus image; importing the same fundus image into a microhemangioma lesion recognition model, a hemorrhage lesion recognition model and an exudation lesion recognition model for recognition; extracting lesion feature information from the recognition results, and then using a trained support vector machine classifier to classify the extracted lesion feature information to obtain a classification result. The microhemangioma lesion recognition model is obtained by extracting a candidate microhemangioma lesion region in the fundus image and inputting it into a CNN model for training; the hemorrhage lesion recognition model and the exudation lesion recognition model are obtained by labeling a region in the fundus image as a hemorrhage lesion region and an exudation lesion region, and then inputting the result into an FCN model for training. The present invention reduces the requirement for the description ability of the network model so as to make the model easy to train, and can locate and outline the lesion regions for different lesions so as to facilitate clinical screening.