Methodology for structuring a dataset of cervical cell images for the study of malignancy changes associated in conventional Pap Test

Abstract

The Pap test is a screening procedure used worldwide for the diagnosis of cervical cancer. The MalignancyAssociated Changes (MAC) are slight morphological and textural changes that take place in the chromatin ofthe cell nucleus. Their study allows the early detection of several types of cancers. However, the number ofdata set for MAC studies is limited by the technical aspects that must be followed. In this work, we createda data set for the study of MAC in images of cervical cells from conventional cytology exams. To extract thecharacteristics of the nuclei, the color channels were separated by the contribution of Hematoxylin and Orange G6dye, which identified morphological, illumination and texture characteristics. As a result, 1186 nuclei of patientseither with some type of anomaly or without anomalies were obtained. To separate the regions of the nuclei, thenon-linear regression function and theWatershedalgorithm with an accuracy of 85.6 % were used. This dataset contains adequate data to complement future work in MAC study of cervical cells from conventional pap smears.