Application of Artificial Neural Networks for Prognostic Modeling in Lung Cancer after Combining Radiomic and Clinical Features

This research paper aimed to investigate the use of machine learning (ML) and artificial neural networks (ANNs) in prognostic modeling for lung cancer, utilizing high-dimensional data. The study utilized a computed tomography (CT) dataset of 422 inoperable nonsmall cell lung carcinoma (NSCLC) patients with tumor segmentation and survival status. Radiomic data extraction was performed using Computation Environment for Radiation Research (CERR). The survival probability was initially determined based on clinical features alone and then using unsupervised ML methods. Supervised ANN modeling was performed using direct and hybrid modeling approaches, which were compared.

The results showed that survival analyses based on clinical features alone were not significant, except for gender. ML clustering using both radiomic and clinical data demonstrated a significant difference in survival. Direct ANN modeling using multilayer perceptron (MLP) yielded better overall model accuracy compared to radial basis function (RBF). Hybrid modeling, combining MLP with feature selection using ML, resulted in an overall model accuracy of 80%. There was no significant difference in model accuracy between direct and hybrid modeling approaches.

In conclusion, this preliminary study supports the application of ANN in predicting outcomes based on radiomic and clinical data for lung cancer patients.