Diabetic retinopathy image labeling method based on deep learning

Abstract

The invention discloses a diabetic retinopathy image labeling method based on deep learning. The method comprises the following steps: selecting a training sample and a test sample; wherein the samplepreprocessing comprises the steps of cutting, overturning and normalization processing of a sample image; constructing a deep full convolutional neural network as an image encoder, using a transfer learning method to initialize network parameters, and then sending the preprocessed sample image data into the network to obtain a feature vector; and constructing a decoder taking a deep recurrent neural network LSTM as a vector, sending the obtained feature vector into the LSTM structure for decoding, and obtaining the annotation information of the sample image. The annotation information is explanation of the focus point information in the sample image, so that a doctor and a patient can be helped to understand the image content more deeply, and the diagnosis efficiency and precision are improved. According to the method, the diabetic retinopathy image is automatically labeled by using the deep convolutional neural network and the deep recurrent neural network, so that the comprehensiveness of focus point information in the image is improved.

Description

technical field [0001] The invention relates to the technical field of medical image processing, in particular to a method for labeling diabetic retinopathy images based on deep learning. Background technique [0002] At present, diabetes is an endocrine disease that seriously affects human health. It is estimated that the number of patients will increase to 380 million by 2025, and its morbidity and mortality are second only to cardiovascular and cerebrovascular diseases and cancer. Diabetic retinopathy (DR) is a common retinal complication associated with diabetes that seriously affects vision and has the highest incidence. In addition, a well-known challenge of DR is that it has no obvious clinical diagnosis in the early stage, making its diagnosis a huge challenge, and patients often do not find out that they have DR until the later stage of the disease. According to statistics, DR has gradually become the main cause of blindness in adults aged 20-74, especially in deve...

AI Technical Summary

Problems solved by technology

There are also many experts using deep learning methods in DR diagnosis to achieve good results, but these algorithms all only classify DR lesions and do not provide diagnostic results in a way that is easy for patients and doctors to understand

Images