Automated Classification of Cervical Image Based on Deep Neural Network
Mengying Zhao, Liyan Zhang, Juan Wang, Chengyi Xia · 2022 IEEE 11th Data Driven Control and Learning Systems Conference (DDCLS) · 2022
Cervical cancer is one of the most common gynecological malignancies. The colposcope can observe the cervical surface through the microscopic biopsy, which can enhance the diagnosis rate of the cervical cancer and help patients to receive timely and effective treatment. Here, this study combines the deep learning and traditional machine learning algorithm to make the classification of the cervix. Firstly, a series of preprocessing operations are carried out on cervical images, which can improve the accuracy and efficiency of the subsequent cervical classification. Secondly, this study uses the traditional machine learning algorithms and the classical neural network to verify the classification performance of cervical datasets. Thirdly, AlexNet-SVM model uses AlexNet convolutional neural network (CNN) to extract the features of cervical images at the front end, and then input the extracted feature parameters into the support vector machine (SVM) classifier at the back end. Meanwhile, the proposed AlexNet-SVM model utilizes the transfer learning method of freezing convolution layers of the model to improve the classification accuracy. The experimental results show that the accuracy of AlexNet-SVM model is 91.77%, the precision is 93.72%, the sensitivity is 86.15% and the specificity is 95.83%. The current work will be helpful to classify and detect the early cervical lesions.