Author: J Isabelle Choi, Meng Wei Ho, Isabel Lin, Hang Qi, Qi Zeng, Ajay Zheng 👨🔬
Affiliation: New York Proton Center, Scarsdale High School 🌍
Purpose: Proton therapy is a promising treatment modality for breast cancer due to its unique dosimetric advantage. Accurate dose calculation is critical to ensure effective treatment outcomes. The following study investigates the dose calculation accuracy for breast cancer patients with very thin chest walls for the Pencil-Beam Convolution Superposition analytic model (Eclipse_PCS) and two Monte Carlo (MC) algorithms (Raystation (RC_MC) and Eclipse_AcurosPT).
Methods: Proton plans with two fields (0°,30°) were generated in Raystation (V2023B) using RC_MC to deliver a uniform 250 cGy dose to a 2.5 cm water-equivalent thickness target sandwiched by air. Measurements at depths of 3 mm, 10 mm, and 20 mm were compared to planned dose distributions calculated by RS_MC, Eclipse_PCS, and Eclipse_AcurosPT. Additionally, treatment plans for 13 post-mastectomy patients being treated to the chest wall and regional lymph node basins were forward calculated using these algorithms. The thin chest wall target was segmented into two regions: skin (first 3 mm) and chest wall target (CWT). Dose metrics of the mean dose to skin and CWT and D95% of CWT were compared to evaluate algorithm limitations in handling the unique air-tissue-lung heterogeneity.
Results: Three algorithms demonstrated good consistency on CWT metrics. The RS_MC presented the highest dosimetric consistency for all measurements and was used as the reference for skin dose comparison. The 13 cases studies showed that Eclipse_PCS always overestimated skin doses by 2.4%±0.7%, while the Eclipse_AcurosPT algorithm underestimated 1.3%±0.6% compared to RS_MC. 2D dose comparisons of three algorithms against measurements at all depths passed >90% in Gamma analysis (2%/2mm/10% threshold).
Conclusion: The RS_MC offers accurate dose calculation for thin chest wall targets. In contrast, Eclipse Eclipse_PCS and Eclipse_AcurosPT algorithms often exhibit limitations, particularly surface dose estimation. Clinicians should account for potential underestimation or overestimation when using these algorithms to prevent target under-coverage or unexpected toxicities.