Convolutional neural networks for predicting molecular profiles of non-small cell lung cancer

Quantitative imaging biomarkers identification has become a powerful tool for predictive diagnosis given increasingly available clinical imaging data. In parallel, molecular profiles have been well documented in non-small cell lung cancers (NSCLCs). However, there has been limited studies on leveraging the two major sources for improving lung cancer computer-aided diagnosis. In this paper, we investigate the problem of predicting molecular profiles with CT imaging arrays in NSCLC. In particular, we formulate a discriminative convolutional neural network to learn deep features for predicting epidermal growth factor receptor (EGFR) mutation states that are associated with cancer cell growth. We evaluated our approach on two independent datasets including a discovery set with 595 patients (Datset1) and a validation set with 89 patients (Dataset2). Extensive experimental results demonstrated that the learned CNN-based features are effective in predicting EGFR mutation states (AUC=0.828, ACC=76.16%) on Dataset1, and it further demonstrated generalized predictive performance (AUC=0.668, ACC=67.55%) on Dataset2.