Radiomic Feature Selection Using Gradient Loss of Deep Neural Network for Lung Cancer Stage Detection
Researchers have developed a new framework called Gradient-Loss Recursive Feature Elimination (GL-RFE) to improve the selection of radiomic features for lung cancer stage detection. This method uses a deep neural network's gradient sensitivity analysis to identify the most impactful features from high-dimensional medical imaging data. The GL-RFE framework successfully identified a top set of 15 features, which were then used to train a classifier achieving over 90% accuracy in distinguishing between early and advanced lung cancer stages. AI
IMPACT Enhances AI's role in medical diagnostics by improving feature selection for high-dimensional imaging data.