Research On Label-free Cell Detection Technology Based On Two-dimensional Light Scattering And Deep Learning

Cells are the basis of all life activities and one of the basic research objects in the fields of biology and medicine.Cell detection technology can be divided into labeled detection and non-labeled detection.Compared with the detection of labeled cells,the label-free detection technology does not require operations such as fluorescent labeling and staining of cells and has the advantages of fast detection speed,non-invasiveness to cells and simple operation.Light scattering technology is an important method for label-free detection of cells.Light scattering occurs when cells or small particles are encountered during light propagation and the scattered light can reflect the relevant information of the cells or particles.There are complex organelles,nucleus and cytoplasm inside the cell.These scattered light images contain rich information such as the size,shape and internal structure of the cell.By analyzing the image characteristics of these scattered lights,cell label-free detection can be achieved.This paper proposes a label-free cell detection technology based on two-dimensional(2D)light scattering and deep learning.The main works of this project are as follows:(1)Through the analysis and research of the current light scattering image method,this topic proposes a 2D light scattering image analysis of cells based on deep learning,so as to realize the label-free detection of cells.Compared with the traditional single image feature extraction algorithm,it can extract richer image features and improve the reliability of cell classification.(2)Completing the construction of a 2D light scattering experimental platform which includes an image excitation module,an image acquisition module,and an image analysis module.Collecting 2D light scattering images of different samples such as red blood cells,white blood cells,polystyrene microspheres,etc.It's to verify the reliability of this experimental platform.(3)The method based on 2D light scattering and deep learning proposed in this paper is used to realize the classification of macrophage subtypes.In this project,2D lightscattering images of cells were obtained by using a built-up 2D light-scattering platform,and related data sets were produced.Using YOLOV3 network to train and verify the data set produced,the sensitivity of the result is 87.8%,the specificity is86.67%,the accuracy rate is 87.21%,and the AUC value reaches 0.9053.(4)The method based on the 2D light scattering and deep learning method proposed in this paper,the label-free detection of human colon cancer cells with different activities is realized.The YOLOV3 network was used to classify and detect the 2D light scattering images of human colon cancer cells and compared with Faster-RCNN and SSD networks.YOLOV3 network has better performance in accuracy,specificity,sensitivity and speed.The result has a sensitivity of 93.6%,a specificity of 94.4% and an accuracy of 94%,with an AUC value of 0.9826.The above results show that the method based on two-dimensional light scattering and deep learning proposed in this paper can effectively realize the label-free detection of cells.