Covid-19 chest CT image recognition method, device and electronic equipment

Abstract

This application relates to a COVID-19 chest CT image recognition method, device and electronic equipment. The method obtains the chest CT image of COVID-19, and constructs a new coronary pneumonia CT recognition network for the characteristics of the chest CT image, trains the network to obtain a COVID-19 chest CT image recognition model, and uses the model to test the CT image sort. Use hole convolution, depth convolution and point convolution operators to reduce redundant parameters; use parallel structure connection to achieve multi-scale feature fusion and reduce model complexity; use downsampling and use maximum fuzzy pooling to reduce aliasing Effect, maintain the translation invariance of the signal; use the channel shuffling operation, reduce the amount of parameters and calculations, improve the classification accuracy, introduce the coordinate attention mechanism, make the spatial coordinate information and channel information be paid attention to, suppress unimportant information, and Solve the resource matching problem.

Description

technical field [0001] This application relates to the technical field of image recognition, in particular to a COVID-19 chest CT image recognition method, device and electronic equipment. Background technique [0002] At present, the main medical methods for diagnosing pneumonia caused by the new coronavirus infection (COVID-19) include chest computed tomography (chest CT), chest X-ray image detection, and magnetic resonance imaging (MRI). For COVID-19 radiological examination, volumetric CT scan is the first choice, with a scan slice thickness of 5 mm (16 slices or more can be achieved), and the reconstruction is a thin slice of 1.0-1.5 mm. Based on thin-slice CT reconstruction, observation in the transverse, sagittal, and coronal planes is conducive to early detection of lesions, assessment of the nature and extent of lesions, and discovery of subtle changes that are difficult to observe with direct digital radiography (DR). However, for radiologists, reviewing lesion in...

AI Technical Summary

Problems solved by technology

However, the existing recognition methods have a complex network structure, a large amount of parameters, and a large amount of calculation. The unimportant information of the chest CT image has a negative impact on the classification results.

Images