Quantitative assessment of deformable image registration accuracy for cumulative dose estimation in multi-site stereotactic re-irradiation

Re-irradiation of previously treated anatomical sites requires precise cumulative dose estimation to balance tumor control with organ-at-risk (OAR) tolerance. Conventional physical dose summation fails to account for inter-treatment anatomical deformations, potentially leading to systematic over- or underestimation of OAR doses. This study evaluates intensity-based deformable image registration (DIR) for improving dose accumulation accuracy in multiple stereotactic ablative radiotherapy (SABR) re-irradiation across different anatomical sites. Sixty patients (226 lesions) who underwent robotic CyberKnife-based SABR re-irradiation (2–5 courses) to liver (36.7%), brain (36.7%), lung (10%), and spine (16.7%) metastases between 2016 and 2025 were analyzed. Cumulative OAR doses were calculated using both DIR-based approach and traditional physical summation. DIR registration accuracy was quantified using dice similarity coefficient (DSC), jaccard index (JI). Biologically effective dose (BED) and equivalent dose in 2 Gy fractions (EQD2) conversions were evaluated for dose assessments across heterogeneous fractionation schedules. DIR-based dose accumulation demonstrated significantly lower maximum doses compared with physical summation: liver maximum dose decreased from 123.8 ± 55.6 Gy to 83.2 ± 28.0 Gy ( p < 0.001) and chest wall maximum dose reduced from 82.9 ± 22.4 Gy to 66.9 ± 18.6 Gy ( p = 0.05). DIR yielded excellent spatial agreement for relatively rigid structures, including the heart, liver, kidneys, and brainstem (DSC > 0.90; JI > 0.84), whereas highly deformable organs such as the bowel and stomach demonstrated markedly inferior agreement (DSC < 0.50; JI < 0.35). Biological dose assessment revealed that DIR-guided workflow more accurately reflected cumulative dose patterns. Deformable registration improves cumulative dose estimation accuracy for multiple SABR re-irradiation. Integration of validated DIR with quantitative quality assurance and biological dose modeling provides a robust framework for multiple re-irradiation workflows. Future prospective studies correlating DIR-derived doses with clinical toxicity are warranted to validate the clinical impact of improved dosimetric accuracy.