KEY POINTS
- The retrospective study included 60 patients with 226 lesions receiving 2–5 CyberKnife SABR/SRS courses between 2016 and 2025. Treated regions included liver (36.7%), brain (36.7%), spine (16.7%), and lung (10%).
- Previous planning CTs and dose distributions were registered directly to the most recent CT using six-degree-of-freedom rigid alignment followed by intensity-based B-spline deformable registration. Physical dose, BED, and EQD2 accumulation were evaluated.
- Estimated liver maximum dose fell from 123.8 ± 55.6 Gy with physical summation to 83.2 ± 28.0 Gy with DIR (p < 0.001), a mean difference of approximately 40.6 Gy.
- Chest-wall maximum dose similarly decreased from 82.9 ± 22.4 to 66.9 ± 18.6 Gy (p = 0.05), approximately a 19% reduction.
- Relatively rigid or consistently shaped structures—including heart, liver, kidneys, and brainstem—showed strong geometric overlap, generally with Dice coefficients >0.90 and Jaccard indices >0.84.
- Performance deteriorated markedly in highly deformable gastrointestinal structures. Bowel and stomach Dice coefficients were below 0.50 and Jaccard indices below 0.35, making voxel-level accumulated dose substantially less trustworthy in those organs.
- The study had no deformable phantom, in-vivo measurement, or clinical toxicity endpoint providing independent ground truth. Consequently, the large differences between methods demonstrate reconstruction sensitivity rather than proving that DIR produced the true accumulated dose.
CLINICAL TAKEAWAY
Simple arithmetic addition of previous dose distributions can be misleading when anatomy changes substantially between treatment courses. DIR is potentially valuable for re-irradiation, but its confidence must be organ-specific; bowel and stomach accumulation in particular should not be accepted without rigorous review and uncertainty management.