Application of artificial neural network and multiple linear regression models for predicting survival time of patients with non-small cell cancer using multiple prognostic factors including FDG-PET measurements

We hypothesize and demonstrate that artificial neural networks (ANN) can perform better than multiple linear regression models in overcoming the limitations of the current TNM staging system for predicting the overall survival time of patients with non-small cell lung cancer (NSCLC). Better prognostication of survival was achieved by including additional prognostic factors, such as FDG-PET measurements and other clinical and pathological prognostic factors. The use of an ANN resulted in a substantial improvement in correlation between actual and predicted months of survival in 328 patients with NSCLC. The ANN resulted in an increase in R², from 0.66 to 0.774, and a reduction in standard deviation, from 17.4 months to 14 months, when compared to multiple linear regressions. Furthermore, the cross-validation results of R²=0.608 suggests that the ANN model was capable of predicting survival for patients who were not included in the database for building the ANN model.