To assess the reliability and target positioning reproducibility with eyes closed in uveal melanoma patients treated with a micromultileaf-based linear accelerator dedicated for stereotactic radiotherapy. Five consecutive patients treated with curative radiotherapy for uveal melanoma were monitored for positioning reproducibility with resimulation CT scans performed every two days while on treatment (23 resimulation CTs available). All patients underwent MRIs of the orbits before simulation to help to define the target and organs at risk (e.g., lenses, optic nerves, ciliary bodies, and lacrimal glands) in the simulation CT (MRI-to-CT bone registration). Patients were simulated, resimulated, and treated with eyes closed. Patient #1 was treated with 5 daily fractions while patients #2 to #5, were treated with 10 daily fractions. We chose the lens of the tumor-bearing eye as the structure to be controlled, assuming that correct repositioning of the lens should be a valid surrogate for correctness of target repositioning. Displacements (mean and standard deviations, SD) of the lens in the three axes were measured for each patient. Systematic and standard errors were calculated. Planning target volume (PTV) margins were estimated according to McKenzie et al. [Phys Med Biol 45, 3331-3342 (2000)].

For both AP-PA and left-right shifts calculated SD were always below 1 mm, except for patient #4, who was treated with a non-customized bolus that pushed the globe backwards in a random fashion. In ideal set-up conditions PTV margins around the target were estimated to be 3 mm. Asking patients to close their eyes is a simple and reliable immobilization procedure when treating ocular tumors with stereotactic radiotherapy. Margins of 3 mm around the target may be necessary to safely treat these tumors under ideal set-up conditions.

Key words: Uveal melanoma; Stereotactic radiotherapy; Organ motion; Micro-multileaf collimator; Novalis shaped beam surgery.