Neural network modelling of proton RBE values at predominant survival fractions of in vitro data

AI Insight

This study applies neural network modelling to predict the relative biological effectiveness (RBE) of protons across predominant survival fractions derived from in vitro experimental data. The models were trained to capture the complex, non-linear relationships between physical beam parameters and biological response, offering a data-driven alternative to fixed RBE values currently used in clinical proton therapy. The findings suggest that machine learning approaches can improve RBE predictions by accounting for variability that deterministic models tend to oversimplify.

Why it matters

Proton therapy is increasingly used to treat cancer with high precision, and more accurate RBE modelling could reduce uncertainty in dose prescription, potentially improving tumour control while limiting damage to surrounding healthy tissue.