Application Of Object Detection Algorithm In BreastPathological Image Analysis

Cancer is one of the most common malignant tumors in the world,and the incidence of cancer is rising in recent years,which seriously affects the physical and mental health of patients.Pathological diagnosis is an important means of diagnosis for cancer,served as "gold standard" of clinical diagnosis.But there is a great shortage of pathologists in China,and a lot of pathological analysis depends on these few doctors,which causes long-term overloaded work to pathological doctors,at the same time,it delays the valuable treatment time for cancer patients.In recent years,the popularization of digital pathology technology and the rapid development of deep learning make the realization ofpathological computer-aided diagnosis possible.Pathologicalcomputer-aided diagnosis has great potential to improve the efficiency and accuracy of pathological diagnosis of cancer.In order to explore the possibility of constructing pathologicalcomputer-aided diagnosis system by deep learning,twokey technologies of breast cancer pathological image analysis were studied in this paper,that is automated detection of nucleus and mitotic cells.The main work and key technologies used in this paper are summarizedbelow:First,the previous research work in the field of pathological image analysis was summarized,and comprehensive research wasdonetocatchupwith the latest progress of deep learning.Second,in order to tackle with the specific task of nucleus and mitotic cell detection,this paper made use of RCNN(Region-based Convolutional Neural Networks)and made a lot of improvements,and these methods have proved to be useful and achieved good scores on the open data set.Finally,further study is needed in this field hoping that more progress can be made.