Intelligent Detection for CT Image of COVID-19 using Deep Learning

Abstract

Purpose: Use artificial intelligence technology to identify the characteristics of covid-19 in CT images, quickly screen COVID-19 patients, achieve rapid diversion and treatment of suspected patients, reduce the risk of infection and control the spread of the disease.Materials and methods: This article combines deep learning target detection and image classification methods to study the CT images of COVID-19 patients. By extracting and analyzing the features of lesions in different periods, a new detection model of covid-19 based on time-spatial sequence convolution is obtained. The algorithm is based on a recurrent neural network structure and a 2D convolutional layer structure.Results: The spatiotemporal sequence convolution kernel based on time and space attributes can effectively extract the latent image semantic features of multiple image data of COVID-19 patients. By comparing with Faster RCNN, YOLO3 and SSD algorithm models, the detection method proposed in this paper can obtain more accurate comprehensive detection results.Conclusion: The time-spatial sequence convolution model can quickly complete the automatic detection of COVID-19 and improve the efficiency of preliminary diagnosis. By correlating images from different stages of the same patient, more accurate auxiliary preliminary screening results can be obtained.