Prediction of high-risk human papillomavirus after conization by machine learning methods

Objectives This study aimed to use machine learning, a branch of artificial intelligence, to predict the persistence of high-risk HPV in women who have undergone conization surgery. Materials and Methods: This retrospective study was conducted between 2018 and 2023 in the Gynecology and Obstetrics Clinic of Balıkesir University Health Practice and Research Hospital. A dataset of 69 female patients between the ages of 23-67 years; for the prediction of HPV status 1 year after the conization operation, the patients' data were recorded according to the criteria we determined, and these data were analyzed and classified using machine learning methods. Various Machine Learning methods such as Gradient Boosting, Support Vector Machine (SVM), Catboost, Random Forest (RF), and Naive Bayes (NB) are used here. Results: We found the highest accuracy rate in Random Forest, and Catboost with 76 %. Gradient Boosting followed with a score of 67%, and Naive Bayes and Support Vector Machine (SVM) performed considerably lower, with scores of 48% and 43%, respectively. Conclusions: Our results show that machine learning, a novel use of artificial intelligence, is effective in predicting the persistence of high-risk HPV. Further studies with more data will be a promising and useful tool for HPV and cervical cancer screening in the future.