Research On Leukocyte Automatic Recognition Algorithm Technology In Leucorrhea Microscopic Image

The automatic identification algorithm of white blood cells in microscopic leukorrhea images is a digital image processing technology closely related to machine learning methods and artificial intelligence.The presence of vaginal leukocytes is an important marker of inflammation in the vagina or cervix.The identification and counting of white blood cells is a simple and effective means of assessing the condition.And it is an important part of routine examination of leucorrhea.However,the identification of white blood cells is mainly performed by manual methods,and the efficiency is low.So far,there have been intelligent automatic identification of the components in the feces,blood and other substances,but the automatic identification of white blood cells in leucorrhea has just started.The realization of auto-recognition relies on the automatic leucorrhea tester.It draws on and uses image processing,feature extraction,intelligent classification,neural network and other algorithmic techniques.It mainly includes two stages: microscopic leukorrhea image preprocessing,and the intelligent classification of white blood cell samples.The pretreatment of the microscopic leukorrhea image completed the segmentation and cutting of the microscopic leukorrhea image and obtained positive and negative sample libraries of leukocytes,which is the basis and the first step to realize the automatic identification of leukocytes.After that,three kinds of algorithm models for the classification of white blood cell samples are proposed in this paper.Extract the morphological features and texture features of the samples,form feature sets and input them into the classifier for classification.Design an algorithm model using convolutional neural network.Feature extraction is performed from different levels and layers,and the classification steps are integrated into the network.The above two algorithm models are linearly combined to obtain a combined algorithm model.After the classification of white blood cell samples,the positive samples are framed and counted in the microscopic leukorrhea image.The experimental results show that the white blood cell recognition methods based on these three different algorithm models can be used in actual production.The method based on manual design features and classification achieved 88% sensitivity,97% specificity,and 92.5% accuracy.The method based on the convolutional neural network obtained 95% sensitivity,89% specificity and 91% accuracy.Finally,93% sensitivity,96% specificity,and 95% accuracy were achieved based on the combined model approach.It can be seen that these three methods have advantages and disadvantages under different evaluation indicators,but the difference is very small and can be used for automatic identification of white blood cells in microscopic leukorrhea images.Considering the time of the program running,the algorithm model based on the convolution neural network presents the best performance.In this paper,from the perspective of algorithm,the automatic identification technology of white blood cells in microscopic leukorrhea images is studied.According to the specific recognition scenarios,manual feature set and convolutional neural network are designed,and the two are combined creatively.The automatic identification of white blood cells under a variety of algorithms is achieved,meeting production requirements and increasing the intelligence of the leucorrhea detector.