Early detection of cervical cancer using deep neural networks

Abstract

Cervical cancer is a disease that is mostly preventable, but it is one of the major causes of cancer fatality in women worldwide. Several studies say that annually 2,60,000 women die because of cervical cancer. Chronic infections with ”high-risk (HR)” human papillomavirus are the leading cause of cervical cancer (HPV). Regular cervical cancer screening, on the other hand, can help to prevent this dangerous disease. Cervical cancer screening is a procedure for detecting precancerous and cancer in women who are at risk, and it is recommended for all women aged 30 to 49. Cervical cancer can be avoided if precancerous lesions are detected and treated early. Nowadays, several tests are performed to detect cervical cancer, most of whom are time consuming and expensive. In this paper, we are approaching the development of a fast and effective system to detect cervical cancer from the cervix image in a minimum time with better accuracy using deep neural networks. First, we collected image data and classified them using VGG16, VGG19, InceptionV3, ResNet50 and ResNet101. From our result we got an accuracy rate of 88.48% from VGG16, 88.97% from VGG19, 88.09% from InceptionV3, 88.67% from ResNet50 and 89.06% from ResNet101. Then, using a mixture of classifiers with the greatest accuracy, we created ensemble models with the best overall accuracy rate of 94.20 percent for CERVIXEN V1, 95.01 percent for CERVIXEN V2, and 94.69 percent for CERVIXEN V3.