Quantitative analysis of motion-induced dose uncertainty and robust optimization in liver lattice radiotherapy: a comparative study

Abstract

Background The precise delivery of high doses to lattice vertices in liver lattice radiotherapy (LRT) can be compromised by respiratory motion.This study aimed to quantitatively characterize the dosimetric impact of respiratory motion on liver LRT and to assess the potential of four-dimensional (4D) robust optimization in mitigating motion-induced uncertainties. Methods Twenty-one patients with bulky liver tumors underwent 4D computed tomography scanning. Three lattice target volumes (LTVs) were generated using varied lattice geometries defined by vertex sizes and spacing. Conventional and 4D robust plans were generated on average intensity projection images. Four-dimensional dose distributions were reconstructed via deformable image registration-based dose accumulation across all respiratory phases. Dosimetric parameters of LTVs and organs at risk (OARs) were compared between static three-dimensional (3D) and dynamic 4D scenarios. Spearman correlation coefficients were calculated to quantify the relationship between motion amplitudes and dosimetric variations. Results Respiratory motion introduced substantial dose degradation in conventional plans. Specifically, LTV Dmin and D98, which represent the minimum dose and the dose covering 98% of the structure, decreased by up to 12.71% and 9.45%, respectively. Motion amplitude was strongly correlated with dose deviations (r>0.7). In contrast, 4D robust plans significantly improved dose stability, with discrepancies in Dmin and D98 between 3D and 4D dose metrics reduced to within 1.4%. Robust optimization led to a significant decrease in the conformity index (CI) in 3D scenarios, but the CI of these robust plans significantly improved in 4D evaluations (P<0.01). Furthermore, under 4D evaluation, robust plans maintained clinically acceptable target conformity, with no statistically significant differences observed compared to conventional plans in most configurations. OAR doses remained within clinically acceptable limits for both conventional and robust plans. Conclusions Liver LRT is highly sensitive to respiratory motion, resulting in notable target dose deterioration under conventional planning. Four-dimensional robust optimization demonstrated the ability to address motion-induced uncertainties and facilitate robust target coverage while maintaining favorable conformity in 4D scenarios through necessary high-dose volume expansion. These findings highlight the importance of integrating motion-aware optimization and patient-specific motion management strategies in clinical LRT planning.