期刊名称：IAENG International Journal of Computer Science
印刷版ISSN：1819-656X
电子版ISSN：1819-9224
出版年度：2021
卷号：48
期号：3
语种：English
出版社：IAENG - International Association of Engineers
摘要：This paper is concern about segmentation on pap smear images to use for classification cervical cancer. A cervical cancer screening by recognizing the cell shape pattern on the Pap-smear image can provide information about the presence of cervical cancer Pap-smear. The manual screening process for classifying cells is a challenging endeavor prone to the risk of error. There are several studies on the classification of cervical cell images which are sometimes not segmented first. Segmentation is very important to get the features contained in the image of cervical cells including the cell nucleus and cytoplasm. However, the segmentation results are also determined by the quality of images. This study proposed 2 paths that had a combination of image segmentation and classification. Before segmentation, image enhancement was carried out to improve image quality using Normalization, CLAHE and Adaptive Gamma Correction. The first path, the segmentation based on the CNN architecture. The second path is a classification process to test the segmentation results used by applying the KNN and ANN methods. The accuracy (ACC) result at the segmentation path was used to measure the match between the location of the segmented pixels and the results on ground truth, while the accuracy at the classification path was the success of the machine in classifying or predicting the segmentation results based on the class label for each group according normal and abnormal Pap-smear’s label. The ACC result obtained in segmentation was only 0.77. However, the results of the segmentation when used in classification to classify which Pap-smears were normal and which are abnormal could provide excellent accuracy results, which were above 0.9. The results of the performance of Sensitivity (SN),Specificity (SP) and F1-score on the segmentation path also only gave results below 0.72, but the performance on the classification path that used the results on the segmentation path gave well results, which were above 0.85. These results stated that the proposed method is quite good in detecting cervical cancer disorders based on Pap-smear images.
