Dosimetric impact of intrafractional ocular motion in LINAC-based ocular SRS/SRT: Defining real-time eye-tracking...

Photon-based LINAC ocular stereotactic radiosurgery and stereotactic radiotherapy (SRS/SRT) offer a noninvasive treatment approach for ocular malignancies. However, intrafractional ocular motion can compromise target coverage because of steep dose gradients and small treatment margins. Although ocular position monitoring and motion-management techniques have been investigated previously, quantitative dosimetric guidance relating angular gaze deviation to target and OAR degradation in modern LINAC ocular SRS/SRT remains limited.

To characterize the dosimetric effects of simulated gaze deviation in photon-based LINAC ocular SRS/SRT, identify planning and geometric factors associated with motion robustness, and derive dosimetrically informed gaze-deviation thresholds for real-time eye tracking.

Ninety-eight previously generated photon-based ocular SRS/SRT plans were retrospectively analyzed for ocular motion robustness using an Eclipse scripting approach. Planning target volumes (PTVs) were derived from hybrid CT-MRI and synthetic datasets based on clinically treated plaque brachytherapy cases, with a median PTV of 2.37 cc in the primary cohort. An ocular motion model was developed to simulate gaze deviations from 0° to 30° across 36 sampled directions per deviation level while incorporating both rotational and translational globe motion. All plans were prescribed 25 Gy in one fraction. A total of 109, 368 simulated dose evaluations were analyzed in the primary D95-normalized cohort, with corresponding PTV D98-normalized plans used to evaluate the effect of prescription normalization. Additional paired subgroup analyses evaluated beam energy (6 MV FFF vs. 10 MV FFF), MLC configuration (standard MLC vs micro-MLC), and PTV margin size (2 vs. 1 mm). The primary robustness endpoint was maximum safe gaze deviation, defined using preservation of simulated V100 ≥ 90%. Additional analyses evaluated OAR doses and associations between baseline geometric and dosimetric variables and motion robustness.

Target coverage progressively deteriorated with increasing simulated gaze deviation. Using V100 ≥ 90% as the prespecified target-coverage criterion, at least 75% of D95- and D98-normalized plans maintained acceptable worst-direction coverage through approximately 6° and 9°, respectively. Optic nerve D0.03 cm³ increased with gaze deviation, with the 75th percentile remaining at or below 12.1 Gy through approximately 7° for D95-normalized plans and 6° for D98-normalized plans. Lens D0.03 cm³ exceeded the 5-Gy reference threshold in most plans at baseline and was therefore not used to determine a motion threshold. Greater target distance from the globe center strongly predicted reduced motion robustness, whereas larger target volume was associated with greater robustness. Higher plan conformity was also associated with increased motion sensitivity. Beam energy, MLC configuration, PTV margin, and tumor directionality had less practical influence on the derived robustness thresholds.

Photon-based LINAC ocular SRS/SRT demonstrated progressive target-coverage loss and increasing selected OAR doses with increasing simulated gaze deviation. Considering both target coverage and optic nerve dose, an approximately 5° gaze deviation represents a reasonable dosimetrically informed starting point for prospective real-time ocular motion monitoring. Because robustness varied substantially among plans, this value should not be considered a universal clinical threshold, and plan-specific assessment may ultimately support more accurate individualized gaze tolerances.