Medical image intelligent diagnosis method based on multi-network integration

Abstract

The invention discloses a medical image intelligent diagnosis method based on multi-network integration, and the method comprises a training stage and a verification test stage, and the training stagecomprises the steps: carrying out the preprocessing of a collected original esophageal cancer data set, and dividing the preprocessed data into a training set, a verification set and a test set; loading a plurality of sub-models, training the sub-models, and taking the trained sub-models as sub-input ends of the integrated model; connecting the output ends of the sub-models by adopting a connection function, combining the output of each sub-model, then setting a single-layer full connection layer or a double-layer full connection layer, and finally performing integrating to obtain an integrated model; preparing related parameters to train the integrated model to obtain a trained integrated model; performing a verification test on the trained integrated model to obtain a classification result; experimental results verify the effectiveness of the Stacking technology in neural network integration, and the recognition rate and generalization performance can be improved through network integration.

Description

technical field [0001] The invention relates to the technical field of intelligent diagnosis, in particular to a method for intelligent diagnosis of medical images based on multi-network integration. Background technique [0002] Cancer is a disease caused by epithelial cell mutation, loss of self-regulation ability, and massive proliferation. It is also called malignant tumor. Malignant tumors are already one of the high incidences in China. In 2018, there were 4.2 million new cases of cancer in China, and the mortality rate was as high as 66.8%, accounting for about 23.7% and 30.0% of the global cancer incidence and death in 2018. pose a serious threat. The clinical manifestations of malignant tumors are also very different due to the difference in the organ, location and incidence of cancer. In addition, malignant tumors often have no obvious symptoms in the early stage, and even if there are symptoms, they are often non-specific. During the diagnosis process, doctors ...

AI Technical Summary

Problems solved by technology

[0004] Purpose of the invention: In order to overcome the deficiencies of the prior art, the present invention provides a medical image intelligent diagnosis method based on multi-network integration, which solves the problem of low recognition accuracy of a single network in the absence of data sets for certain diseases

Images