Research Of Cell Image Segmentation And Recognition Based On Deep Learning

Cell image segmentation and recognition is an important part of medical image processing. Medical cell images have different sizes, uneven image color and easy adhesion between cells. At present, the existing cell image segmentation algorithms are not ideal for the segmentation of the cell images which contour are not clear and have the low color contrast. Therefore, getting accurate segmentation results from the cell images are very difficult. At present,deep learning is a hot issue in machine learning research. The purpose of deep learning is to establish and simulate the neural network of the human brain to analyze and study. It can mimic the human brain's way of thinking to interpret data and extract data information features. In view of the great advantage of depth learning technology in image processing, this paper applies it to cell image segmentation and recognition.The main research work of this paper is as follows: 1) based on the convolutional neural network, combined with simple linear iterative clustering algorithm, a new algorithm model for cell image segmentation is proposed.First of all, the cell images are preprocessed using staining correction method,to improve the color contrast of the image; then using convolutional neural network to obtain preliminary segmentation results; finally send the simple linear iterative clustering boundary information feedback to the initial image segmentation on the improved lifting. The proposed algorithm can effectively reduce the redundancy of the image local information, and obtain the boundary position of the target area more accurately. Experimental results show that the proposed method is better than the classical convolutional neural network and other cell segmentation algorithms, such as the OTSU and WATERSHED method. Through quantitative cell segmentation results analysis, the segmentation accuracy reached 92.72% in the Breast Cancer Cells dataset. 2)based on the study of cell image segmentation, we propose a method based on convolution neural network and the method of ZCA whitening treatment. The cell images used the method of staining correction and ZCA whitening treatment has a higher contrast and lower correlation between the data.Therefore, it is more suitable for research and analysis. Compared with the experimental results, it is proved that the proposed method is superior to some traditional image recognition and classification methods, such as SVM and Softmax. The accuracy of the improved cell image recognition method reached 95.67% in the breast pathology cell images dataset.