2022-RA-757-ESGO Bioimpedance screening for cervical cancer. (20th October 2022)

• Record Type: Journal Article
• Authors: Zhang, Tingting; Heo, Jun Beom; Mok, Jinwon; Oh, Tong In; Yeo, Seung Geun; Hong, Ji Yun; Park, Dong Choon
• Is Part Of: International journal of gynecological cancer. Volume 32(2022)Supplement 2
• Page Start: A72
• Page End: A72
• Publication Date: 2022-10-20
• Subjects: Generative organs, Female -- Cancer -- Periodicals; 616.99465
• DOI: 10.1136/ijgc-2022-ESGO.159
• Languages: English
• ISSNs: 1048-891X

Abstract

Abstract : Introduction/Background: The pathological changes at different severity stages of cervical intraepithelial neoplasia (CIN) can give rise to the differences in electrical properties of the cervical tissues, which lay the foundation for screening of CIN using the non-invasive and cost-effective technique of electrical bioimpedance as well as paves the way of easy availability of CIN screening, especially in developing countries and thereby the reduction of the occurrence of cervical cancer. Methodology: The complex impedance measurements collected by a multi-electrode probe from the clinically diagnosed 117 normal subjects, 18 subjects with the low-grade squamous intraepithelial lesion (LSIL), and 19 subjects with the high-grade squamous intraepithelial lesion (HSIL) have been fitted to a Cole-Cole model, which is used to characterize the electrical properties of biological tissues. The geometrical and electrical features have been extracted from the model and later exploited to classify the pathological conditions of cervical tissues using the receiver operating characteristic (ROC) curve based on three algorithms: logistic regression, support vector machine, and Naive Bayes. Results: Regardless of the algorithms being used, the area under curve (AUC) obtained from the features of complex impedance is apparently higher than that of using the real part of impedance only, and the ROC areas between normal and LSIL are much more improved when complex features are exploited, as shown in figure 2. Furthermore, due to the significant discrepancy between the electrical properties of normal and HSIL cervix, greater AUC can be observed in the separation of normal and HSIL cervix, irrespective of the features and algorithms being used. Conclusion: Compared with the utilization of only the real part of impedance measurements, our results present superior performance when using the combination of geometrical and electrical features from complex impedance in improving the effectiveness of classification during the screening of CIN according to the ROC analysis.