Research On Pathological Test Of Cervical Epithelia With Multi-source Pathological Images And Machine Learning

With the growth of national population and the changes in living conditions,the needs of healthcare resources are increasing.In recent years,the combination between artificial intelligence and medical science has gained more and more attention in academia,and it also becomes the new trend of pathology.Compared with the traditional pathological examination,the intelligent methods not only can improve the accuracy,but also reduce the workload of the pathologists,which can enhance the efficiency.Therefore,it is very important to create an intelligent model of pathological examination,which is stable and reliable.This paper proposed a novel model to test the carcinomatous transformation in cervical epithelium.This model analyses three kinds of pathological images,which are widely used in the test of cervical epithelium.They are hematoxylin-eosin staining,immunohistochemistry-P16 and immunohistochemistryKi67.This model is able to improve the accuracy and efficiency of pathological examination and give the integrated diagnostic results.The content of this paper are:1.This paper analyses the development and tendency of digital pathology and illustrates the significance of the combination between artificial intelligence and medical science,and also analyses the relating technology of digital pathology.2.This paper cooperated with Xinqiao Hospital of Chinese Army Medical University to research the pathological section of cervical carcinoma,and analyses the feature of pathological image in pathological examination.3.The proposed model has completed the pathological examination in cervical epithelium by combining three different kinds of features,which are from three different kinds of pathological image.Firstly,this model uses SIFT to capture and match the potential regions from three kinds of pathological image.Secondly,By using SLIC and U-net neural network,this model separates the cervical epithelium from the image and enhances some features.At last,based on residual network and Support Vector Machine,this model obtain the result by analyzing three kinds of image with enhanced features.4.The pathologists from Xinqiao hospital collected plenty of digital pathological image of cervical.Based on these data,this paper carried out the experiment and analyses the experimental results.The experimental results show that the proposed model is very efficient and accurate in cervical epithelium testing,which can provide reliable results for pathologists.