A Study on Various Machine Learning Techniques Used For Colorectal Cancer Disease Prediction and Survival

Abstract

Cancer is a major burden of disease worldwide. Each year, tens of millions of people are diagnosed with cancer around the world, and more than half of the patients eventually die from itCancer is a disease caused by the uncontrolled division of abnormal cells in parts of the body. There are several types of cancer in the world. Inthepast, the diagnosis of cancer mainly depended on the experience and knowledge of the doctor. Butbecausein thepastten years,computerized decision support systems haveplayed a vital role in the healthcare industry. Machine learning methods are used forthe prevention and early detection of cancer patients. The rapid development of machinelearningtechnology does help cliniciansmakecorrect diagnosis decisions. Tumor prediction at the TNM (tumor, nodule, and metastasis) stage of colon cancer has been studied using the most influential histopathological parameters and five-yeardisease-free survival (DFS) predictionthrough machine learning (ML) clinical research.4021 patients wereselectedfor analysis from the ColonCancerRegistry (CRC)at Chang Gung Memorial Hospital, Linkou, Taiwan.Several ML algorithms were usedto predict the tumor stage of colon cancer takinginto account thetumoraggressionscore(TAS).The performance of the various ML algorithms was assessed using five-foldcross-validation, which resultedin an efficient validation of the precision achieved by the algorithms, whichtook intoaccount boththe standard TNM staging cases andthe TNM staging with the tumoraggressionscoreThe random forestmodel was observed to achieve an F-value of 0.89 when tumor aggression was included as the attribute score along with the standard attributes normally used for TNM stage prediction. We also found that the Random Forest algorithm outperformed all other algorithms, with an accuracy of about 84% and an area under the curve (AUC) of 0.82 ± 0.10,around the five-year DFS.