Cervical cancer is the fourth most common cancer among women in the world.There are about 528,000 new cases each year,and about 85% of new cases occur in less developed countries.In these countries,the high mortality rate is mainly due to the lack of skilled medical personnel and appropriate medical pre-screening procedures.With the rapid development of computer science technology and Internet of Things technology,image technology based on automated detection and deep learning is also increasingly used in the process of medical diagnosis and screening of patients’ diseased cells.The onset of cervical cancer is not a sudden process.After HPV virus invades the human body,it will slowly erode the epithelial cells of the cervix,and the cell morphology of the lesions during this process will continue to change.In the early stage of the relevant detection methods,the discovery of abnormal cells and the corresponding treatment will greatly reduce the probability of female death due to cervical cancer.The current detection methods are mostly for doctors to manually carry out cervical cell collection smears and staining observations.Different doctors have different proficiency in their operations,and the time,diagnosis rate and misdiagnosis rate of the entire process are also different.At the same time,repeated operations such as mechanized smearing and staining for a long time may also reduce the attention of the doctor and there is a risk of misdiagnosis.This paper proposes a deep learning-based cervix for the problems of low labor efficiency,high manual participation,high professional requirements in operation,and different efficiency of doctors during the screening process.Cell automatic identification system.The system can assist the doctors in the production and staining process of the collected cervical cells,and after the image is acquired,the cervical cells can be classified and detected,and the cervical cells can be classified into normal cells and diseased cells according to whether the lesions occur,To complete the pre-screening of the cells,but also can provide advice for the doctor whether to conduct in-depth follow-up testing of the patient.The content of the research work in this article is as follows:1.Construction of automation platform.The automation platform is mainly composed of a mechanical motion module and a fluid control module.The mechanical motion module is composed of a motion module of an electric gripper and a sampling needle motion module.In addition,the fluid control module includes a liquid sampling module and a sampling needle cleaning module.Each module realizes the operation of automatic sampling and sampling of slides and test tubes through corresponding sensors.At the same time,the RS485 communication module is used as a tool for communication between the computer and the lower computer hardware module,so that we can control the automation platform through the computer.To achieve our goal of automatic film production and pasteurization.2.Use deep learning algorithms to complete the detection of cervical cells and the identification of cell types.Based on the published data set,the data samples were manually calibrated and the data enhancement of small sample image data,and the YOLOv3 algorithm was used to train the cervical cell data samples,and the cervical squamous Epithelial cell samples are identified to determine whether a lesion has occurred,and finally a variety of evaluation methods are used to analyze the accuracy of the judgment results.This paper combines the preparation and staining of traditional cervical cell smears to realize the operation of assisting physicians in preparing and staining through an automated platform,and the detection model trained by the YOLOv3 algorithm used can effectively treat the cervix Cell detection and classification identification.The purpose of assisting the physician to analyze the collected cell samples is achieved. |