An automatic grading method based on label coding for imaging lesions of the sugar reticulum

Abstract

The invention discloses an automatic grading method for sugar reticulum image lesions based on label coding, wherein the label coding technology can not only be used for the grading problem of sugar network image lesions, but also can be applied to other ordered classification problems. The present invention firstly performs soft coding on the five lesion grade labels of the sugar reticulum lesions to control the model prediction probability distribution, so as to solve the problem of class dependence between the labels; and then builds the sugar network data set and trains the model to obtain model parameters and The offset parameter in the parameter soft label; the soft coding of the five lesion grades calculated by the offset parameter can be used for transfer learning; finally, the model obtained by training can accurately classify the lesions of the sugar network. This method can effectively solve the dependency between labels in the grading of glycemic reticulum lesions. The soft coding method can flexibly control the label prediction probability distribution of the model and improve the accuracy of grading of glycated reticulum lesions. At the same time, hard and soft labels can be established. Mapping for transfer learning.

Description

technical field [0001] The invention relates to the application fields of image processing and machine vision, in particular to the severity assessment and intelligent classification of diabetic retinopathy. The label coding technology in the invention is also applicable to other ordered classification problems in deep learning. Background technique [0002] In the grading of the severity of diabetic retinopathy, according to the International Clinical Diabetic Retinopathy (DR) Severity Scale as shown in Table 1, the lesion grades can be divided into five categories: normal, mild, moderate, severe, and proliferative. Figure 1 shown. Part of the classification of the severity of the lesion belongs to the ordered classification problem, and there is a dependency between the categories. For a certain label, different categories play different roles in predicting the probability distribution. For example, different predictions for value-added (DR4) fundus photographs such as ...

AI Technical Summary

Problems solved by technology

The disadvantage of this algorithm is that it can only convert hard labels into fixed soft label distributions, and does not have generalization ability.

Images