COVID-19 detection method based on collaborative deep learning and lung CT image

Abstract

The invention discloses a COVID-19 detection method based on collaborative deep learning and lung CT images, a model used in the method is composed of two pre-trained residual neural networks (ResNet-50) and one collaborative network, each ResNet-50 learns image representation and classification, the learned image representation is used as the input of the collaborative network, and the collaborative network is used for detecting the lung CT images. The network has a fully connected structure to predict whether input image pairs belong to the same class. Therefore, under the supervision of the classification error of each ResNet-50 and the collaboration error of the two ResNet-50, the collaboration network model can be trained in an end-to-end manner, the network learning feature representation capability is further enhanced, and the easily confused samples can be effectively and accurately discriminated.

Description

technical field [0001] The invention belongs to the technical field of medical image classification and detection, in particular to a COVID-19 detection method based on collaborative deep learning and lung CT images. Background technique [0002] At present, nucleic acid testing is used as the detection result of the new crown. Nucleic acid testing is also regarded as the gold standard for the detection of this disease. However, nucleic acid testing is time-consuming and has certain false negatives. Usually, multiple tests are required to ensure that the test results are correct. In addition, nucleic acid shows certain shortcomings in the detection of early lesions and asymptomatic patients; compared with CT, it has a faster detection speed, and has a wider range of due and cheaper prices. However, due to the great similarity between CT of COVID-19 and pneumonia caused by other viruses, manual clinical diagnosis is time-consuming and labor-intensive, and there is a possibili...

AI Technical Summary

Problems solved by technology

During the global spread of COVID-19, in order to quickly complete its diagnosis and minimize the damage of COVID-19 to the global economic life, many researchers focused on using deep learning detection methods, most of them proposed chest CT scan Methods of analysis, specifically mentioning the problem of differentiating between COVID-19 and other lung diseases, namely pneumonia (CAP) and interstitial lung disease (IDL); Leading to poor accuracy and insufficient generalization ability, coupled with strong inter-class similarities and intra-class differences between them, it brings great trouble to classification and diagnosis problems

[0004] Aiming at the problem of insufficient data, a large number of experiments have proved that the use of transfer learning can effectively solve the problem of small sample learning in medical image classification, but it cannot improve the problem of similarity between classes and differences within classes in medical images.

Images