A Decision Support System for Improving Lung Cancer Prediction Based on ANN

Abstract

Recent advancements in artificial intelligence (AI) and big data analysis have shown great potential for improving the diagnosis of lung cancer. Early detection of lung cancer is crucial for increasing patient survival rates. This paper analyze the data BRFSS (Behavioral Risk Factor Surveillance System), conducted from 2017 to 2020 to identify risk factors and symptoms of lung cancer. We develop a decision support system (DSS) based on data mining technique to assist healthcare practitioners and users in early diagnosis of lung cancer. Thirteen risk factors and demographic data are selected as predictors. The ANN and a logistic regression (LR) model are performed to predict the probability of lung cancer and to serve as a prognostic index respectively. The ANN model shown an accuracy of 84.79%, a sensitivity of 79.8%, and a specificity of 89.76%, a 93% of the ROC (AUROC) curve. While the LR model obtained an accuracy of 80.2%, a sensitivity of 80%, and a specificity of 72.2%, with a 76.1% AUROC. The models are trained with a batch size of 100, using stochastic gradient descent (SGD) optimizer. By using data analysis and mining techniques, we discovered new patterns in the health behavioral risk data that are previously unknown. Overall, our proposed method has a potential to significantly improve the early detection and treatment of lung cancer.