Integrating functional brain regions into treatment optimization: a multi-institutional dosimetric study in CyberKnife SRT for brain metastases

Preserving brain function is becoming an important goal in stereotactic radiotherapy, with increasing attention to the potential impact of radiation dose on functional neuroanatomy. This study investigated whether including these functional brain regions as optimization objectives during SRT for brain metastases can achieve measurable dose reductions. A retrospective analysis was conducted on 41 patients with single brain metastases treated with CyberKnife SRT across two institutions. Patients received 22 Gy in one fraction or 30 Gy in three fractions, according to lesion size and location. For each case, the clinical plan (TP CL ) was generated according to international guidelines. Functional regions within 3 cm of the target were automatically segmented using BrainLab’s multimodal MRI workflow. A function-sparing plan (TP FS ) was then re-optimized using the same target and OAR criteria while incorporating functional regions as avoidance structures. For each structure, D mean and D max were extracted; for small-volume regions (< 0.5 cc), only D max was evaluated. Relative dose variation (ΔD) from TP CL to TP FS was calculated, and differences were assessed using the Wilcoxon signed-rank test. Automatic segmentation identified 29 functional regions across the cohort, ranging from large cortical gyri to small subcortical and allocortical nuclei. TP FS preserved full target coverage, with variations in VDpre < 0.5% and no significant differences relative to TP CL ( p > 0.05). Standard OAR dose metrics and plan complexity remained within clinical tolerance. Across functional regions, TP FS produced a consistent and significant dose-lowering effect. Median ΔD mean ranged from − 22.8% (putamen) to − 39.0% (caudatus), with notable decreases also observed in the hippocampus (− 27.7%), thalamus (− 30.5%), and globus pallidus (− 38.8%). ΔD max showed similar trends, including − 32.0% (thalamus), − 44.6% (globus pallidus), and − 30.1% (hippocampus). A few regions did not show significant ΔD changes: the cerebellum and amygdala for both ΔD mean and ΔD max , the corpus callosum for ΔD mean only, and the postcentral gyrus, thalamic anterior nucleus, and trigeminal nerve for ΔD max ( p > 0.05). Incorporating functional neuroanatomical structures into SRT optimization is technically feasible and leads to consistent dose reductions across multiple cortical and subcortical regions, without compromising target coverage, OAR sparing, or plan deliverability. Although dose–response thresholds for these structures are not yet established, these findings provide a preliminary dosimetric basis for function-guided planning and provide the dosimetric prerequisite for future prospective trials with neurocognitive endpoints.