Three-dimensional knowledge diagnosis model construction method and system

Abstract

The invention discloses a three-dimensional knowledge diagnosis model construction method and system. The method comprises the steps of establishing a training set based on known medical image data, wherein the training set comprises two parts, Xm represents mth medical image data, Ym represents a label of the mth medical image data, Xi represents ith label-free medical image data, and M and N arethe number of corresponding samples respectively; training the constructed three-dimensional knowledge diagnosis model based on the neural network by taking a set loss function as a target, wherein Xm and Ym are taken as input and labels of supervised training of a main branch of the three-dimensional knowledge diagnosis model, Xi is taken as input of unsupervised training of the main branch of the three-dimensional knowledge diagnosis model, and the input of unsupervised training of an auxiliary branch of the three-dimensional knowledge diagnosis model is data obtained by carrying out various disturbances on Xi. According to the invention, efficient and intelligent diagnosis of medical images can be realized, and the method is used for clinical indication.

Description

technical field [0001] The present invention relates to the technical field of medical image processing, and more specifically, to a method and system for constructing a three-dimensional knowledge diagnosis model. Background technique [0002] In recent years, with the rapid development of deep learning, domestic and foreign scholars have used computers to simulate doctors' diagnosis of diseases, and used deep learning to extract and diagnose intelligent diagnostic indicators of diseases. At present, relevant scientific research results have reached the expert level of disease diagnosis, and even outperformed experienced clinicians in some aspects. Although deep learning-based disease diagnosis techniques have made great achievements, these deep learning-based diagnostic methods rely on a large number of medical imaging data with real label samples, and it is very difficult to obtain a large number of clinical medical image data with real label samples . The most common w...

AI Technical Summary

Problems solved by technology

[0005] 2) Due to the quality of the acquired data and the limited real label data, the accuracy of the probability value of the network model to predict the sample data to generate the real standard is not high

[0006] 3), the current commonly used network architecture is based on two-dimensional convolutional neural network, and two-dimensional convolutional neural network does not make full use of spatial information

Images