Artificial Intelligence and Machine Learning for Women’s Health Issues - Chapter 15 - Deadly cancer of cervix tackled with early diagnosis using machine learning

Cervical cancer is a severe malignancy whose initial indications are difficult to recognize. It is the fourth most prevalent cancer in women worldwide, with 604,000 new cases were observed in the year 2020. Of the predicted 342,000 cervical cancer fatalities in 2020, 90% are predicted to occur in low- and middle-income countries. Early identification and prompt treatment can help to avoid this. Cervical cancer screening and early diagnosis using artificial intelligence (AI) and machine learning (ML) are both cost-effective and computationally cheap. ML methods such as decision trees, logistic regression, support vector machines, K-nearest neighbor algorithms, adaptive boosting, and gradient boosting algorithms are used in the prediction of cancer. These algorithms are applied mostly to pap smear images of cervical region or microscopic biopsy images. The strength of the database (e.g., UCI Machine Learning Repository) upon which decision support systems are built has a substantial bearing on how effective they are. Missing values, duplicated characteristics, and unequal class distribution are frequently seen in medical datasets. Prior to using data mining techniques, it is necessary to properly analyze the data to create a viable prediction model. In this study, we used a dataset on cervical cancer risk factors that includes 32 risk variables for cervical cancer in addition to the Schiller, Hinselmann, biopsy, and cytology target variables. These four target variables are the most often utilized vertical cancer diagnostic tests. Accuracies, sensitivities, specificities, positive predictive accuracy (PPA), and negative predictive accuracy (NPA) are used to assess each algorithm’s efficacy. This study demonstrates how ML algorithms might help a medical practitioner identify cervical cancer early and perhaps save lives.