Fundus image detection method, device and system based on machine learning

Abstract

The invention discloses a fundus image detection method, device and system based on machine learning, wherein the method includes the following steps: acquiring a fundus image to be detected; performing a first feature set detection on the entire area of ​​the fundus image; The specific area in the image is detected by the second feature set, and the salience of the features in the first feature is greater than the salience of the features in the second feature set; the conclusion of the detection is judged based on machine learning to obtain the final detection result. Each type of feature is detected separately and does not affect each other. The category of each feature can be judged more accurately, and multiple types of features with multiple salient degrees can be detected at the same time, which can efficiently and accurately detect fundus images.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a fundus image detection method, device and system based on machine learning. Background technique [0002] In recent years, machine learning has been widely used in the medical field, especially machine learning technology represented by deep learning has attracted widespread attention in the field of medical imaging. For example, in terms of fundus image detection, deep learning technology can more accurately detect a certain feature of the fundus image, such as using a large number of samples of the macular hole feature to train the deep learning model, and use the trained model to perform macular detection on the fundus image. Hole detection. These techniques are often limited to the detection of a single feature or a small number of associated features, and cannot accurately detect other features. However, since the eye is a very delicate and complex organ...

AI Technical Summary

Problems solved by technology

Therefore, the use of existing detection technology often leads to the difficulty of convergence of the detection results, resulting in inaccurate detection results

Or, training a model for each feature to detect separately, not only requires a huge number of samples, but also leads to a sharp increase in the amount of calculation when the number of features is large, which in turn leads to a decrease in detection efficiency

Images