Application of Convolutional Neural Network in Raman Spectral Recognition of Covid-19

Abstract

The outbreak of COVID-19 has lasted for two years. The rapid spread and fatal variability of COVID-19 pose a great threat to human survival. Today, the existing high-tech medical technology has not found a direct specific drug. Therefore, efficient diagnostic techniques and methods play a key role in controlling the spread of COVID-19 and managing patients' conditions. Deep learning technology can learn implicit samples of data. This paper mainly studies the nonlinear relationship between the serum Raman spectrum data of new crown and healthy people by using convolutional neural network, effectively expand the samples of training set by using data enhancement method, standardize the spectral data, smooth denoising by savitzky Golay method, and construct the prediction model based on convolutional neural network after principal component analysis. Compared with other traditional machine learning algorithms, the features extracted by convolution neural network through convolution layer, batch standardization layer and pooling layer are more comprehensive, which can effectively improve the accuracy and speed of COVID-19 recognition and classification. The experimental results show that convolution neural network has a higher screening accuracy for COVID-19, and the accuracy rate is 98.39%, It is proved that Raman spectroscopy combined with deep learning is effective and feasible in screening COVID-19.