Research On Cervical Cell Feature Selection And Classification Algorithm Based On Computer Vision

According to the World Health Organization,cervical cancer is the second most prevalent cancer in women worldwide,with more than 500,000 women worldwide suffering from cervical cancer in 2020 and a mortality rate of about 68.4%.The incidence of cervical cancer is high,but early screening can greatly reduce the mortality rate caused by the disease.The most common screening method for cervical cancer is Pap smear,but this screening method relies on manual reading,which is labor intensive and the results are not reproducible.The application of artificial intelligence technology to computer-aided classification of cervical cells can effectively solve these problems.In this paper,we focus on several key technologies for computer-aided classification of cervical cells: cervical cell feature selection,feature-based cervical cell machine learning recognition method,and image-based cervical cell deep learning classification method.The main research of this paper is as follows:(1)Before using machine learning methods for cell classification,feature selection is required.In this paper,we propose a hybrid feature selection method based on filter-wrapper for cervical cell images with many feature dimensions,a large number of redundant features and irrelevant features,and a two-stage processing of the data set.Pearson coefficient was used to analyze the correlation of cervical image features,and the optimal feature subset was screened by the sequential backward selection method wrapped with three classifiers commonly used in machine learning.The experimental results show that the feature selection method proposed in this paper can effectively reduce the required feature dimensionality for classification on all three different datasets.(2)The classification performance of Support Vector Machine(SVM)based on feature classification is affected by its parameters.In this paper,a GWO-SVM model is established to classify cervical cells by using the Grey Wolf Optimizer(GWO)algorithm to optimize the hyperparameters of the Support Vector Machine classifier.Finally,the optimal classification model was used to identify cervical cells.The experimental results show that the recognition accuracy of this model can reach 93%,which is better than the original support vector machine and support vector machine based on grid search,and can better realize the classification of cervical cells.(3)In this paper,ResNet50,an image-based deep learning classification network,is improved by using methods based on transfer learning and label smoothing strategies,and the accuracy rate is 98%.The experiments show that the improved model in this paper is efficient and accurate,and can be further applied to the classification of cervical cells.In addition,the interpretability analysis of the model is carried out to improve the confidence of the model.Both feature-based and image-based cervical cell classification methods can achieve a good level of classification on the cervical cell dataset,and both methods have their own applicable scenarios.The study of feature-based and image-based classification models on the cervical cell dataset SIPaKMeD in this paper can provide an important reference for pathologists interested in machine learning methods and has good application prospects.