COVID-19 early screening and severity degree evaluation method and system based on attention guidance

Abstract

The invention discloses a CT-based attention-guided COVIDI-19 early screening and severity degree evaluation method and system. The method comprises the following steps: preprocessing CT images in early screening and lesion assessment data sets; constructing a COVID-19 early screening model and a severity evaluation model based on CT, wherein each of the COVID-19 early screening model and the severity evaluation model comprises a feature extraction module, a feature fusion module and a classification module; respectively training a COVID-19 early screening model and a severity evaluation modelby utilizing the preprocessed early screening and lesion evaluation data set; preprocessing the test CT image in the step 1, then inputting the preprocessed test CT image into the trained COVID-19 early screening model, and if a new crown is detected, inputting the preprocessed test CT image into the trained severity assessment model to assess the severity of the new crown. According to the method, the complexity of directly learning two tasks by the model is reduced to a certain extent, the problem of small available data volume of medical images is relieved, and the precision of early screening and severe evaluation is further improved.

Description

technical field [0001] The invention belongs to the technical field of medical image processing, and in particular relates to a method and system for early screening and severity assessment of COVID-19 based on CT-based attention guidance. Background technique [0002] COVID-19 has several characteristics that make it difficult to control: 1) COVID-19 is highly contagious and can be transmitted through direct transmission, aerosol transmission and contact transmission. The transmission power of a new crown patient is between 3 and 4. Facing the rapid growth of new crown patients, the medical system of any country will be under tremendous pressure. 2) It is almost impossible for doctors to distinguish patients infected with COVID-19 from interstitial pneumonia caused by influenza virus, mycoplasma, chlamydia, respiratory syncytial virus, bacteria only by clinical features (such as fever, dry cough and dyspnea). And there are many asymptomatic infections. 3) If a cytokine st...

AI Technical Summary

Problems solved by technology

The model has a series of designs for the problem of few samples. (1) First, in order to reduce the complexity and parameter amount of the model, and then reduce the overfitting of the model, we choose to design a 2D network instead of a 3D network for The extraction of CT image features, (2) In order to reduce the demand of the model for training samples in the process of designing the 2D network, we adopted the idea of ​​transfer learning and designed a 2D network that can make full use of pre-trained weights to reduce the need for samples. Requirements (3) Use the designed weight-sharing 2D network to slide and extract the features of CT slices. The design of weight sharing also reduces the parameters of the network and reduces the over-fitting phenomenon of the model.

Images