Comparison of setup accuracy of three different thermoplastic masks for the treatment of brain and head and neck tumors

Purpose: Setup accuracy is an important factor influencing the definition of the planning target volume (PTV). The purpose of this study was to compare the setup accuracy of three different thermoplastic masks used for immobilization of patients with brain or head and neck tumors. Materials and methods: Thirty patients with brain or head and neck tumors were consecutively assigned to one of three different thermoplastic masks (Posifix®): head mask with three fixation points (3 FP, ten patients), head and shoulder mask with four fixation points (4 FP, ten patients), head and shoulder mask with five fixation points (5 FP, four fixations plus an additional one on the top of head, ten patients). Once a week, during the session with a 6 MV linac (Elekta®), orthogonal (antero-posterior and lateral) portal images were acquired for three fictitious isocenters placed during the simulation at the level of the head, the neck and the shoulders. Portal images and digitized simulator films were compared using the PIPS pro software, and displacements in antero-posterior (A-P), cranio-caudal (C-C) and medio-lateral (M-L) directions were calculated. From these displacements, 2D or 3D errors were also calculated. Results: A total of 915 portal images were obtained, of which 98% could be analyzed. For the whole population, total displacements reached a standard deviation (SD) of 2.2 mm at the level of the head and the neck. Systematic and random displacements were in the same order of magnitude and reached a SD of 1.8 mm. Patient setup was slightly worse at the shoulder level with a total displacement of 2.8 mm (1 SD) for both the C-C and the M-L directions. There again, the systematic and the random components were in the same order of magnitude below 2.4 mm (±SD). For isocenters in the head and in the neck, there was no substantial difference in the setup deviation between the three masks. The setup reproducibility was found to be significantly worse (P=0.01) at the level of the shoulders with the 3 FP mask. For the 2D random error, 1 SD of 2.3 mm was observed compared to 0.8 and 1.2 mm for the 4 and 5 FP masks, respectively. Lastly, 90% of the 3D total deviations were below 4.5 mm for the head and the neck. In the shoulder region, 90% of the 2D total deviations were below 5.5 mm. Conclusion: Thermoplastic masks provide an accurate patient immobilization. At the shoulder level, setup variations are reduced when 4 or 5 FP masks are used. These data could be used for the assessment of margins for the PTV.