A data-driven modelling based approach to evaluating prognostic value of Electrical Impedance Spectroscopy for cervical cancer diagnosis

Abstract

Abstract Electrical impedance spectroscopy (EIS) has been used as an adjunct to colposcopy for cervical cancer diagnosis for many years. The study presented in this paper was a longitudinal EIS data analysis where women with a negative colposcopy were followed up to three years and their initial EIS readings were analyzed to see if it was possible to predict the women who subsequently developed cervical cancer. A data-driven modelling approach was proposed to extract features from EIS readings and cross validation techniques were then used to choose the best classification model constructed from the selected features to separate the group of women who developed cervical cancer from those who didn’t within the follow-up years. The developed method was applied to analyze a real EIS data set and the results showed that EIS does offer prognostic information on the risk of cervical cancer development over three follow-up years. The method developed is of long-term benefit for EIS-based cervical cancer diagnosis and, in conjunction with standard colposcopy, there is potential with the developed method to provide more effective and efficient patient management strategies for clinic practice.