Geometry-based framework for beam angle selection in proton therapy for lung cancer

Background and purpose: Proton therapy is a promising modality for treating locally advanced non-small cell lung cancer (LA-NSCLC). However, respiratory-induced intra-fractional motion can compromise tumour coverage. This work aimed to develop a method to identify proton beam angles that balance tumour coverage and organ-at-risk (OAR) sparing. Materials and methods: An open-source dataset of eleven LA-NSCLC patients with four-dimensional CT (4DCT) scans was analysed. For each gantry-couch angle, the water equivalent path length variation (ΔWEPL) and percentage irradiated volume (PIV) were calculated using in-house code to serve as metrics for tumour motion sensitivity and geometric OAR exposure. A unified risk map was constructed based on ΔWEPL and PIV, with patient-specific weighting factors and constraints to enable individualised beam selection. Subsequent single-beam treatment plans were generated on the average intensity projection CT, with the dose distribution recalculated on all breathing phases to derive ΔD95% (variability in target D95%). Pearson’s correlation tested associations between ΔWEPL and ΔD95% and between PIV and OAR dose. Results: ΔWEPL correlated with target dose degradation (median r = 0.90 for ΔD95%), while PIV correlated with OAR doses (r = 0.88–0.98 across heart, lungs, and spinal cord). Risk maps identified beam angles minimising motion sensitivity while respecting OAR constraints. A three-beam patient example illustrated that clinically acceptable tumour coverage was maintained while substantially reducing lung and heart dose compared to suboptimal beam orientations. Conclusions: Patient-specific proton beam angle selection by combining ΔWEPL and PIV into a unified risk map offers a clinically relevant strategy to improve robustness and OAR sparing for LA-NSCLC.