Machine learning-based prediction of survival and recurrence in patients with colon cancer

Abstract

This article provides an overview of colorectal cancer statistics through the machine learning algorithm to gain insight into survival and mortality rates and relapse trends. In this study, three basic machine learning algorithms were used for prediction: support vector machine (SVM), genetic algorithm, and XGBoost decision tree. Additionally, some basic statistics were also used. By utilizing two different clinical datasets and the histopathology parameters, the prediction of metastasis and survival time of colon cancer was analyzed using machine learning (ML) in clinical research. The first dataset used in the study is the longitudinal dataset that comprises 929 patients and the clinical study results observed for approximately 6.5 years. The second dataset consists of a 90-person data set obtained from patients with APC level II tumors. With these data, the most appropriate model was selected using machine learning methods, and the survival and tumor recurrence predictions were made and evaluated. It was concluded that there was a notable difference in prognosis and a prominent difference in terms of gender between the early- and late-stage relapse groups. It should be emphasized that the most important factor affecting survival time in the study is the time to recurrence. Moreover, it was observed that the time to relapse and the time of death were the same in most of the study.