Study On Near Field Microwave Imaging Of Cells

Observing the complex microstructure of cells is an important step in the study of cells.Modern biology,medicine and even medicine need to explore the state,characteristics and physiological process of biology with the help of the characterization of cell microstructure.Fluorescence labeling is the most representative of traditional cell imaging methods.This kind of method has the advantages of low cost,high sensitivity and simple operation,but the use of fluorescent reagent will damage the cell itself.With the deepening of research,more new requirements are put forward for cell imaging and analysis technology.New high-resolution biological imaging methods are needed to obtain information about the changes of cell properties.As a new micro imaging technology,near-field microwave microscopy system can characterize the electrical parameter response of samples in the region less than the free space radiation wavelength,which can break through the spatial resolution limit of far field microscopy.Because microwave is sensitive to dielectric properties and penetrating,microwave near-field microscope also has great potential in biological imaging.More importantly,the use of microwave imaging does not require additional processing of biological samples,which ensures the nondestructive testing of samples.This paper studies the imaging of plant cells based on the near-field microwave microscope system,and uses the machine learning method to process the data collected by the near-field microwave microscope system for cell classification.The core of the near-field microwave microscopy system is the response between the probe tip and the scanned sample.In this paper,the simulation of the tip sample response is carried out in COMSOL multi physical field simulation software,and the effects of different factors such as cell dielectric characteristics,water content and tip spacing on plant cell scanning imaging are analyzed.We analyzed the problem of image distortion in the scanning process and adjusted the scanning control program of near-field microwave microscope.Through the two imaging of drum chip,the effect of image improvement by control software adjustment is verified.After that,we cultured two kinds of plant cells polluted and unpolluted by heavy metal ions through hydroponic culture.Based on the adjusted control software,the two kinds of onion epidermal cells were scanned by point scanning,line scanning and surface scanning,and the near-field imaging results were obtained.The comprehensive comparison of optical photos,near-field imaging results and point scanning images shows that the near-field microwave microscope system has the potential to realize cell recognition.The scanning data are classified by machine learning algorithm to realize the cell recognition of near-field microwave microscopy system.Firstly,the data are preprocessed according to the scanning characteristics of near-field microwave microscopy system.In addition,in order to solve the problem of data imbalance,the scanned data are expanded with synthetic minority oversampling technique.Then the chi square test and mutual information method are used to score the features of the feature vector,and the eigenvalues with high correlation are selected as the input data.Finally,back propagation neural network,one-dimensional convolutional neural network and integrated model are used for classification prediction,and the performance of different classification methods is compared.The results show that the classification accuracy of the method can reach90.93%.