Segmentation Of Cell Images Based On Convolutional Neural Network And The Discrimination Of Cell Type

With the large-scale implementation and promotion of the latest cutting-edge technologies such as artificial intelligence and big data,the deep learning field based on convolutional neural network has been a hot spot in the scientific research field in recent years.Based on advantages of deep learning in image retrieval classification and feature extraction fusion,the main content of this paper is apply convolutional neural network to the field of medical image segmentation and conduct classification and discrimination research,so as to realize the application of deep learning technology to the analysis and research of medical image data.In view of the different sizes,shapes and textures of medical images,it is difficult to segment accurate cell regions.This paper proposes a new algorithm based on convolutional neural network combined with edge clustering for cell images segmentation.Firstly,the color contrast of the original image samples is improved by the dye correction preprocessing method,and then the preliminary segmentation results are obtained by the convolutional neural network.Finally,the edge clustering is used to improve the continuity and integrity of the preliminary segmentation results.In addition,this paper also use deep learning target detection technology for cell area target detection has also achieved certain effects,more intuitive display of effective target areas in cell images,helping medical scientists identify and determine,and provide objective data reference for pathologists.Experiments show that compared with classical convolutional neural networks,fuzzy clustering,threshold segmentation and other cell image segmentation algorithms,the cell segmentation method proposed in this paper improves the integrity of segmentation results by 6.15%;compared with the classical VGG19 structure enhancement 1.17%.In the actual clinical medical diagnosis process,the size of cells is generally used as a reference for judging the type of cell physiological state.Based on this,the current popular computer vision technology is used to obtain the basic properties of the cell particles in the segmented image,named circumference,area,etc,and cell type discrimination by a support vector machine classifier.In addition,this paper also combines the commonly used methods in the existing machine learning to classify and cluster the cell particles,and also obtains the effect of effectively identifying the classification.