Medical image auxiliary annotation method and system based on block chain

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a medical image auxiliary annotation method and system based on a block chain. The method comprises the following steps: reading a neural network for a target disease on a block chain, extracting high-dimensional features of each piece of image data, and storing the high-dimensional features on the block chain; reading all high-dimensional features on the block chain and clustering the high-dimensional features to obtain a plurality of categories; obtaining a to-be-labeled high-dimensional feature by using a neural network, and calculating the difference between the to-be-labeled high-dimensional feature and the mean feature of each category, and the over-fitting degree and the local fitting degree of each category to obtain the over-fitting suppression capability of the to-be-labeled high-dimensional feature; obtaining the under-fitting suppression capability of the to-be-labeled high-dimensional features; according to the over-fitting suppression capability, the under-fitting suppression capability and the similarity between the to-be-labeled high-dimensional feature and the expected feature, obtaining a labeling priority of the to-be-labeled high-dimensional feature; and labeling the to-be-labeled high-dimensional features in sequence. According to the embodiment of the invention, data more useful for parameter learning of the neural network can be marked, and a data set is expanded.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, in particular to a block chain-based medical image auxiliary labeling method and system. Background technique [0002] With the advancement of medical technology, hospitals have begun to use advanced medical equipment to diagnose diseases, such as X-ray and other perspective imaging equipment, endoscopic imaging equipment or nuclear magnetic resonance equipment, etc., to visualize the patient's condition in the form of image data Come out, assist doctors to diagnose the disease more quickly and accurately and determine the treatment plan. [0003] With the development of artificial intelligence technology, it has begun to use various neural networks to diagnose diseases based on medical image data. The most commonly used technique is to use neural networks to segment medical image data to obtain lesion regions of organs on the image. Since the medical image data of diffe...

AI Technical Summary

Problems solved by technology

[0004] However, since the labeling of medical image data requires professionals, resulting in fewer labeled datasets, even the neural network trained by federated learning will face the problem of fewer datasets. The data set is not shared, and it is impossible to perform data cleaning and other preprocessing operations on the data of all hospitals, which may make the data distribution of all hospitals not wide enough, or the features contained in the data are not comprehensive enough, and only contain some features

These reasons will cause the actual accuracy of the trained neural network to not be too high, and there will be problems of over-fitting and under-fitting

Images