Nuclear Classification Detection And Counting System Based On Deep Learning

The level of cell proliferation is a very important index in both clinical diagnosis and biomedical experiments.According to research,the level of cell proliferation can reflect a lot of pathological information.Through the analysis of cell proliferation level,we can grasp the cell activity state and abnormal state of patients,and then make corresponding judgments on the rehabilitation process or disease development trend of patients,so as to provide help for doctors’ diagnosis.To judge the level of cell proliferation,it is often necessary to calculate the number and proportion of proliferating cell nuclei.Due to the different experience of doctors and the defects of traditional detection schemes or software,when the number of nuclei is large and densely distributed,problems such as fatigue,long detection time and low accuracy will appear.Therefore,automatic detection of proliferating cell nucleus and automatic counting from cell pattern will save a lot of time for doctors to judge and improve the detection accuracy.At this point,this paper proposes a method based on deep learning,which can automatically classify and count all kinds of nuclei with large number and dense distribution,and compare with the current mainstream methods used in hospitals,which effectively solves the problems of long time and low accuracy of manual and semi manual detection.The main contents of this paper are as follows:1.Data acquisition.Rat liver cells were used as the data source of this study.The proliferating cell nucleus was stained by IHC,and the non proliferating cell nucleus was not stained.Then Leica dm3000 system was used to collect pictures.2.Data annotation.Labelimg is used to label the above image data to generate a machine-readable XML file.3.Data enhancement.We use data enhancement algorithm to expand our data set by translating and cutting the data.4.Model training.Using the RetinaNet model,we get several models under different algorithms such as RESNET and Vgg16,verify the advantages and disadvantages of different algorithms,and evaluate the accuracy of the model.5.Contrast experiment.Using the current hospital mainstream method to process the same batch of data,the results are compared with our method.6.Software packaging.Considering the application value of this study,we design an operable UI interface for the model,so that people who do not have relevant algorithm knowledge can simply use our model.The model obtained in this paper realizes the automatic classification and counting of proliferating cell nuclei,and is significantly ahead of the mainstream methods in hospitals in terms of detection accuracy.In terms of detection time,our model can complete the detection task of 30 minutes in seconds.Experiments show that the scheme can effectively help doctors to evaluate the level of cell proliferation,so as to carry out rapid diagnosis.