Background: Frameless immobilization allows for planning and quality assurance of intensity-modulated radiosurgery (IM-SRS) plans. We tested the hypothesis that IM-SRS planning with uniform tissue density corrections results in dose inaccuracy compared to heterogeneity-corrected algorithms.
Methods: Fifteen patients with tumors of the pituitary or cavernous sinus underwent frameless IM-SRS. Treatment planning CT and MRI scans were obtained and fused to delineate the tumor, optic nerves, chiasm, and brainstem. The plan was developed with static gantry IM-SRS fields using a pencil beam (PB), analytical anisotropic (AAA), and Acuros XB (AXB) algorithms. We evaluated measures of target coverage as well as doses to organs at risk (OAR) for each algorithm. We compared the results of each algorithm in the cases where PTV overlapped OAR (n = 10) to cases without overlapping OAR with PTV (n = 5). Utilizing film dosimetry, we measured the dose distribution for each algorithm through a uniform density target to a rando phantom with non-uniform density of air, tissue, and bone.
Results: There was no difference in target coverage measured by DMaxPTV, DMinPTV, D95%PTV, or the isodose surface (IDS) covering 95% of the PTV regardless of algorithm. However, there were differences in dose to OAR. PB predicted higher (p < 0.05) Dmax for the brainstem, chiasm, right optic nerve, and left optic nerve. In cases of PTV overlapping an optic nerve (n = 7), PB was unable to limit dose to 8 Gy while achieving PTV coverage (PB 855 cGy vs. AAA 769 cGy, p = 0.05 vs. AXB 658 cGy, p = 0.03). Within the rando phantom, the PB and AAA algorithms over-estimated the dose delivered in the bone-tissue-air interface of the sinus (+17%), while the AXB algorithm closely predicted the actual dose delivered through the inhomogeneous tissue (+/- 1 % max, p < 0.05).
Conclusions: Patients undergoing frameless SRS benefit from heterogeneity corrected dose plans when the lesion lies in areas of widely varying tissue density and near critical normal structures such as the skull base. Film dosimetry confirms that the AXB dose calculation algorithm more accurately predicts actual dose delivered though tissues of varying densities than PB or AAA dose calculation algorithms.