COVID-19 identification method based on multi-information sample class adaptive classification network

Abstract

The invention relates to a COVID-19 identification method based on a multi-information sample class adaptive classification network, and the method comprises the steps: collecting different classes of human chest X-ray images of different data sources and corresponding personal experience data, and carrying out the classification, and making an original data set; performing data cleaning on the original data set, performing feature enhancement on an X-ray image in the original data set, and extracting a lung region; arranging the lung region data set and the personal experience data to form a target data set; training the multi-information sample class adaptive classification network by using the target data set; and inputting a human chest X-ray image to be recognized and corresponding personal experience data into the trained multi-information sample class adaptive classification network to obtain a COVID-19 recognition result. According to the recognition method disclosed by the invention, the features are extracted from various kinds of information, and the COVID-19 is recognized according to the extracted features, so that the recognition accuracy is improved; and sample class adaptive classification is adopted, so that the adverse effect of class imbalance on training under the condition that the number of samples is small is relieved.

Description

technical field [0001] The invention belongs to the field of medical image classification, and in particular relates to a COVID-19 recognition method based on a multi-information sample class self-adaptive classification network. Background technique [0002] The rapid spread of the novel coronavirus disease (COVID-19) has become a global concern. To curb this spread, in clinical practice, RT-PCR testing is often complemented with chest X-ray imaging (CXR), such that combined analysis can reduce a large number of false negatives while providing more information about the extent and severity of the disease. [0003] In order to find a faster, more objective, accurate and sensitive method for disease identification and assessment, there is currently a research trend that uses clinical features extracted from CXR for automatic detection. [0004] A potential benefit of studying chest X-ray images is that they can characterize pneumonia status even in asymptomatic people. Howe...

AI Technical Summary

Problems solved by technology

[0005] However, the accuracy of the detection results of the existing pneumonia detection methods based on CXR images is not high, the detection results are not convincing, and the accuracy of the detection results is seriously affected when the training samples are lacking

Images