Mitigating dosimetric perturbations from large metal implants utilizing spot-scanning proton arc therapy

Conventional intensity modulated proton therapy (IMPT) techniques are exquisitely sensitive to setup and range uncertainties in the presence of high-density metal implants. Herein we introduce a spot-scanning proton arc (SPArc) approach to improve the nominal plan quality near metal implants by increasing the degrees of freedom. 4-field IMPT and single-arc SPArc plans were generated for 5 patients with target volumes circumscribed or traversed by metal implants. Robust optimization was used (parameters: ± 3.5% range, ± 3/5 mm setup uncertainties). Twenty-six perturbation scenarios per patient were generated to assess robustness. Nominal plan quality was evaluated by dose volume histogram (DVH) target coverage and hotspot metrics. Target volume homogeneity was assessed via standard deviation (SD), interquartile range (IQR), range width, Levene's variance test, and homogeneity index (HI). Robustness was quantified via Levene and mean DVH 95% confidence interval (95% CI) width. Compared to IMPT, SPArc significantly improved target coverage at all tested volume and dose levels. Hotspots were significantly reduced at V_105%. Generalized equivalent uniform dose (gEUD) was significantly greater with SPArc. DVH analysis revealed greater volumes at V84.7% to V100.2% (mean 8.5% volume advantage) and volume reduction from V_104.6% to V_107.5% (mean 4.6% volume reduction). SPArc plans were more homogeneous by SD (2.3 vs 6.2%), IQR (1.9 vs 4.1%), range (22.7 vs 56.2%), HI (mean: 0.07 vs 0.21, p = 0.03), and by Levene (statistic: 65, p < 1 × 10-14). SPArc was more robust to setup uncertainties as measured by Levene (mean statistic: 1.1 vs 1.6) and mean DVH 95% CI width (mean: 1.82 vs 2.52%, p = 0.052). In comparison to IMPT, SPArc therapy achieved substantially better homogeneity, robustness, and target coverage in the immediate presence of large metal artifacts. This provides a novel solution for a challenging clinical scenario.