Prediction of two year survival among patients of non-small cell lung cancer

Abstract

Lung cancer ranks as second most prevalent type of cancer. Still predictions for survival of lung cancer patients are not accurate. In this research, we try to create a prediction model, with the help of machine learning to accurately predict the survival of non-small cell lung cancer patients (NSCLC). Clinical data of 559 patients was taken for training and testing of models. We have developed multilevel perceptron model for survival prediction. Other models developed during this study were compared to measure performance of our model. Attributes that are found to be useful as biomarkers for prediction of survival analysis of NSCLC have also been computed and ranked accordingly for increase in accuracy of prediction model by implementing feature selection method. The final model included T stage, N stage, Modality, World Health Organization Performance status, Cumulative Total Tumor dose, tumor load, Overall treatment time as the variables. Two year survival was chosen as the prediction outcome. Neural Network was found as the best prediction model with area under Curve (AUC) of 0.75. By far to our knowledge Multilevel Neural Network is found to be the best model for predicting two-year survival among patients of non-small cell lung cancer.