Cervical Cancer Image Processing with Convolutional Neural Network for Detection

Abstract

The diagnostic method for detecting cervical cancer using Pap smear can be laborious and time-consuming. Therefore, research on computer-aided diagnosis is essential. The purpose of this study is to aid the distinguishing of Pap smear images from various categories of cervical cells by creating an alternative image processing and classification method. This is so that in the future, the burden on pathologists to manually analyze many Pap smear images can be reduced. The developed method will be able to help in the detection of abnormality or cancer. The processing methods include Gaussian filtering, Otsu thresholding, Canny edge detection, and Convolutional Neural Network. The analytical methods utilized were accuracy and loss curves, and the evaluation measures of accuracy, precision, recall, and F1 measure. The most optimal trained model had an accuracy, precision, recall, and F1 measure of 93.26%, 92.55%, 91.52%, and 91.84% respectively. It was concluded that the image processing and classification method could be used to distinguish multi-cell Pap smear images. Even with some limitations, it has the potential to improve single-cell analysis and also aid in classification. In the future, this method may be used in the medical field to help diagnose cervical cancer in Indonesia.