Optimization of spatiotemporally fractionated radiation therapy for brain cancer patients : Spatiotemporal fractionation schemes for SRS treatment of patients with multiple brain metastases

Background: Stereotactic radiosurgery (SRS) is increasingly preferred to whole-brain radiotherapy for a limited amount of brain metastases, but its extension to patients with many lesions is not without controversy. Concerns regarding distant brain failure, dose spillage and treatment times have put SRS into question as a viable treatment approach. This thesis investigates whether spatiotemporally fractionated (STF) plans, where the dose distribution between fractions may vary, can improve the therapeutic ratio for such patients. A constrained approach to STF plans (cSTF) is also proposed, with the goal of improving robustness. Method: Spatiotemporal fractionation schemes aim at an improvement in therapeutic ratio by achieving some degree hypofractionation in lesions, and hyperfractionation in healthy tissue, by varying the dose distribution in targets with a more uniform fractionation in healthy tissue. Three fraction intensity-modulated plans were generated in RayStation for two representative cases: Patient A (6 metastases) and Patient B (28 metastases). Fluence map optimization was performed directly, with convertion of fluence dose to biologically effective dose (BED). Additionally, a constrained approach to spatiotemporal fractionation schemes forced each fraction to treat the entire lesion in the cSTF formulation, which should improve robustness while preserving the ability to vary the total dose on the lesion. Plans were assessed with mean BED to normal brain, V60BED2 and V60BED2 Gy BED, gradient (GI) and conformity (CI) indices, and D98 to each planning target volume (PTV). Results: For Patient A, STF lowered mean brain BED2 by 9% but increased the volume of healthy tissue receiving high doses, whereas cSTF achieved a comparable 9% BED reduction with only a 3% rise in V84BED2 Gy and achieved similar tumor control, compared to uniformly fractionated plans. For patient B, STF reduced mean BED by 12% with an increase in V84BED2 of 5%. cSTF presented the largest benefit of 19% reduction in mean BED together with 13% and 7% reductions in V60BED2 and V84BED2 , respectively, while maintaining adequate BED10 D98 in almost all targets. Conclusions: Spatiotemporal fractionation schemes can reduce the biological dose delivered to the healthy brain, but the unconstrained form may compromise robustness and increase volumes of healthy tissue receiving higher doses. The cSTF approach retains the reduction of biological dose, yields dose distributions more similar to that of clinical practice, all with no increase in high dose areas of healthy brain and similar tumor control. These findings suggest that cSTF is a promising step toward clinically deliverable SRS plans with spatiotemporal fractionation schemes. Future work should focus on robustness analysis and machine-parameter optimization to investigate and review the demonstrated spatiotemporally fractionated plans as clinically deliverable treatment plans.