Using Deep Learning To Explore The Intelligent Diagnosis Of Lymphatic Invasion

Malignant tumor cell metastasis is the main cause of death in patients with malignant tumors,while lymphatic metastasis is the main way of malignant tumor cell metastasis.Lymphatic Invasion is one of the main manifestations of lymphatic metastasis and an important prognostic indicator for most malignant tumors.The traditional diagnostic method for Lymphatic Invasion is to observe the hematoxylin-eosin(HE)stained pathological slide under a microscope,and to identify and count the Lymphatic Invasion based on the morphological characteristics of Lymphatic Invasion under the microscope.Because pathologists use this method in the daily pathological diagnosis of Lymphatic Invasion and the diagnostic results are not objectively reproducible,pathologists need to do D2-40 immunohistochemical staining for auxiliary diagnosis.Although D2-40 immunohistochemical staining has a low rate of missed diagnosis of Lymphatic Invasion,the cost of immunohistochemical staining is also high.This study intends to use the artificial intelligence deep learning Faster RCNN model to realize the detection of Lymphatic Invasion;use the deep learning Cycle GAN to map the digital pathological image features of HE stained Lymphatic Invasion to D2-40 immunohistochemical stained Lymphatic Invasion pathological image features.Finally,HE Lymphatic Invasion digital pathological images through the Cycle GAN to generate virtual "D2-40-like immunohistochemically stained Lymphatic Invasion " digital pathological images.To explore the pathological diagnosis of Lymphatic Invasion,we only need to make HE stained pathological slide.Then use the D2-40 immunohistochemical staining Lymphatic Invasion deep learning Faster RCNN detection model to detect the structure of the Lymphatic Invasion in the digital pathological image of "D2-40-like immunohistochemical staining Lymphatic Invasion",to help the pathologists realize intelligent quantitative diagnosis of Lymphatic Invasion.Innovative exploration in the diagnosis of Lymphatic Invasion,only HE staining is needed to quickly and accurately complete the quantitative diagnosis of Lymphatic Invasion.This method can not only help pathologists complete the quantitative diagnosis of Lymphatic Invasion efficiently,but also help patients save medical costs for immunohistochemical staining and save social medical costs.In this experiment,the Faster RCNN detection model for Lymphatic Invasion with HE staining is used to detect the Recall value of 0.917 and m AP value of 0.791 for HE staining Lymphatic Invasion.Recall value for detecting D2-40 immunohistochemically stained Lymphatic Invasion is 0.968,m AP value is 0.858;Deep Learning Cycle GAN is used to achieve HE staining Lymphatic Invasion pathological images to D2-40 immunohistochemically staining Lymphatic Invasion pathological images of Lymphatic Invasion are interchanged.The virtual generated "D2-40-like immunohistochemically stained Lymphatic Invasion" pathological image has a SSIM value of 0.0755 and a PSNR value of 12.167;D2-40 immunohistochemical staining of Lymphatic Invasion Faster RCNN detection model is used to detect the Lymphatic Invasion in the pathological image of "D2-40-like immunohistochemically stained Lymphatic Invasion".The Recall value is0.929,and the m AP value is 0.822.This thesis mainly explores the use of artificial intelligence deep learning technology to help pathologists make intelligent diagnosis in the field of pathological diagnosis.The research has realized the use of deep learning Faster RCNN detection model to assist pathologists intelligent detect Lymphatic Invasion in pathological images of D2-40 immunohistochemical staining and HE staining.The deep learning Cycle GAN was used to realize the conversion of pathological images of HE stained of Lymphatic Invasion to D2-40 immunohistochemical staining of pathological images of Lymphatic Invasion,and virtual generation of "D2-40-like immunohistochemistry stained of Lymphatic Invasion" pathological image.The method studied in this paper can provide pathologists with a new method for the diagnosis of Lymphatic Invasion,and change the traditional model of pathological diagnosis of Lymphatic Invasion.At the same time,it can help patients save medical costs for immunohistochemical staining and save social medical costs.Finally,the performance of the deep learning Faster RCNN model and Cycle GAN in this paper still needs to be updated.To be continued in the future to improve the performance of the algorithm,better applied in the field of pathological diagnosis,to provide pathologists with more accurate intelligent pathological diagnosis services.