The Research Of Intelligent Recognition And Analysis Of Cell Morphology In Fully Automatic Blood Smears

Microscopy of blood smears is one of the most important clinical routine tests in hospitals.Usually doctors diagnose the health of the human body by analyzing the proportion of white blood cells in the blood smear.Under current conditions,most hospitals perform blood smear examinations by professional physicians through visual inspection through a microscope.This has high requirements on the quality of professional physicians,and subjective judgments of physicians will cause certain errors in the results of evaluations.Moreover,due to the heavy workload of manual microscopy,effective medical resources are wasted to a certain extent.We completed a set of blood smear intelligent identification and analysis system by studying the white blood cell automatic recognition classification algorithm and related technologies in the blood smear image.The optimization and improvement of the conventional artificial microscopy technology in the hospital can be achieved,which can alleviate the problem of shortage of medical resources.The specific research content of this article is summarized as follows:First of all,this paper designs an automated system based on the image acquisition and performance requirements of the blood smear,and proposes a complete design scheme including hardware,software and algorithm.Secondly,contrary to the need for automatic acquisition of blood smear images,this paper improves on the existing sharpness evaluation function,and proposes a fast focusing method based on different focus positions to ensure the speed of focusing.Then,for the white blood cell segmentation in the blood smear image,this paper proposes a cell nucleus threshold segmentation method based on color space,and an improved Grabcut white blood cell segmentation algorithm based on coordinate positioning.Compared with the traditional segmentation algorithm,the accuracy rate is improved by more than 8%,which effectively solve the problem of inaccurate segmentation in the process of white blood cell segmentation was solved.Next,for the feature extraction and feature selection of white blood cells,this paper separately extracts features from three aspects: morphology,texture and grayscale,and then uses the Filter method to select features,and uses Fisher discriminant thought to filter the features to obtain the final feature set.Finally,for the classification of white blood cells in blood smears,this paper verifies five kinds of white blood cell multi-classification strategies,four of which are based on SVM,and the other is based on the gradient boosting decision tree XGBoost.The five classification methods have advantages and disadvantages for different types of white blood cell classification.After experiment comparison,the system uses the faster and higher accuracy XGBoost method as the classification algorithm.Compared with the traditional blood smear identification method,the automatic blood smear intelligent identification and analysis system proposed in this paper has a faster recognition speed and higher segmentation and classification accuracy.The system accuracy rate reaches 91.12%.In terms of practical use,it has been greatly improved,and has a wider use prospect.