Study On Segmentation Of Urinary Sediment Image And Recognition Of Urine Crystal

The examination of urine sediment image is an important method for clinical examination and diagnosis of urinary system diseases,the traditional way of checking urine sediment imageis through an artificial microscope.This method affects the accuracy of the diagnosis because of the existence of defects,for example,the workload is too large,and it is easy to make subjective errors for a long time,and so on.The traditional diagnostic methods can not be processed or transformed because of the observed images.It is difficult to carry out remote transmission and can not be easily accessed.Urine slag image analysis and processing based on digital image processing technology can greatly improve the efficiency of urine sediment image inspection,reducing the work intensity of the inspectors,and promoting the standardization of medical information and disease diagnosis.This thesis makes a summary of the research work on the image processing of urine sediment and the identification of the types of urine crystals at home and abroad,on the basis of it,the method of dividing the tangible components of urine and identifying the classification of urine crystallization is studied,a set of more complete and effective scheme for image processing and classification is proposed.This thesis mainly focuses on the following aspects:1.In the image preprocessing stage,based on the comparison analysis and experiment of different filtering methods,the mean filter is selected to denoise image filtering.Compared with other denoising methods,it can better protect the tangible component boundary information in the urine sediment image,so that the subsequent segmentation work is easier to carry out.2.The correct segmentation of the image is the key to the accurate analysis of the tangible components.In the phase of image segmentation,according to the complex background of the urine sediment,the characteristics of the composition are complex on the basis of comparison and experiment of different image segmentation methods,in this thesis,a segmentation method based on Prewitt edge detection operator is adopted in the image segmentation of urine sediment.The experimental results show that the edge detection operator based on Prewitt can detect the edges of the visible components of urine sediment accurately,and the segmentation method combined with morphological processing has a good effect on the segmentation of urinary sediment images.3.In this thesis,we apply clustering algorithm to urine sediment image segmentation and extract the urine crystal part.Based on the original K-means algorithm,we propose an improved K-means algorithm based on the characteristics of different components of the urine sediment,the experimental results show that the proposed method can improve the clustering effect more than the original method under the premise of saving time.4.In the classification and recognition of urine crystallization,the AlexNet framework based on convolution neural network is used to identify the urine crystallization,through the experiment,it is found that the classification rate of urine crystallization is higher.In this thesis,we studied the segmentation of urine sediment and the classification and recognition of urinary crystallization.Based on previous works,we integrated our own methods and research ideas,a complete set of urine sediment image segmentation and urine crystallization segmentation method has been initially formed.