Research And Application Of Deep Learning In Cervical Cancer Image Recognition

Cervical cancer,as the most common gynecological malignancy,seriously affects the health of women.The traditional diagnosis method is that the doctor analyzes and judges the patient’s cervical cell smear,local CT,lesion MRI,and other clinical data.However,this method is susceptible to the clinical experience of doctors and subjective judgments of people,leading to misdiagnosis or missed diagnosis.With the rapid progress of artificial intelligence in recent years,the emerging application of assisted diagnosis of diseases has become one of the research hotspots.AI-based diagnosis of disease is mainly divided into clinical data detection methods based on machine learning and image detection methods based on deep learning.The application of this emerging auxiliary diagnostic method in the medical field is gradually mature,such as the detection of clinical features of esophageal cancer,Detection of brain tumor images.However,the research based on artificial intelligence in cervical cancer detection is relatively small,and still faces many difficulties and challenges.First,there are currently no standard MRI dataset used for cervical cancer category detection.Data sets with obvious features and accurate classification can not only enhance the prediction ability of the model,but also improve the convergence during training;secondly,cell smears or CT images are unitary,and the feature information extracted from the image data of the same category after training is quite different,making it difficult to use the model Third,the number of real and effective cervical cancer samples is very limited,and there is also a serious imbalance in the distribution of data categories.Based on the research on the detection of cervical cancer pathological types,this article analyzes and experiments on clinical MRI cases of cervical cancer from a tertiary-grade A class hospital in Sichuan Province,discussed the existing difficulties in depth,and proposed specific solutions.The main contents of this article are as follows:(1)Annotation and establishment of a T2 sagittal image dataset in cervical cancer MRI.This dataset contains a total of 3 categories with a total of 422 images.(2)A data augmentation method based on a deep generative adversarial network was established,which effectively solved the imbalance problem of cervical cancer data samples.(3)An image recognition model based on deep residual neural network is constructed to extract the characteristic information of cervical cancer lesions.Experimental results show that the model is superior to other methods in cervical cancer detection.(4)A clinical feature fusion method based on machine learning was constructed.The experimental results show that the method not only reduces the complexity of the model,but also improves the accuracy of cervical cancer detection.