A Rapid Target-Clustering Metric for Multi-Target Radiotherapy
Abstract
Purpose
Planners must often choose how to cluster multiple SRS/SRT targets without reliable a priori dosimetric prediction, making exhaustive plan-and-optimize testing impractical in complex cases. A previously developed target-clustering metric (M) was designed as a rapid pre-planning tool to score cluster arrangements for expected normal brain/OAR sparing. This study evaluates whether M correlates with delivered dose by comparing extreme-M cluster arrangements.
Methods
A 10-metastasis intracranial stress-test case was created by registering targets from three patients into a single dataset and perturbing lesion positions to generate a challenging but clinically plausible distribution. Candidate target clusters were scored using M (geometric cost incorporating target volume and cluster extent). All four-cluster configurations of the 10 metastases were generated and those yielding the minimum and maximum cumulative M were selected and planned in Varian Eclipse. Each cluster was treated with one isocenter at the cluster center-of-mass. A standardized VMAT SRS template was used per cluster (four arcs: couch 0° full arc; couch 45°, 315°, and 90° half arcs). To control for prescription effects, all targets were treated to 20 Gy × 1 fx, irrespective of size. The resulting multi-isocenter plans were co-optimized across all isocenters and evaluated using Brain-PTV and key OAR dose metrics.
Results
Normal brain V12 Gy was 141.99 cc and 162.69 cc for the minimum-M and maximum-M cluster arrangements, respectively (-12.7%). Corresponding OAR D0.03cc values were 5.87 Gy and 8.71 Gy for the brainstem and 3.44 Gy and 4.14 Gy for the optic chiasm, respectively. Similar reductions were observed in other OARs.
Conclusion
In this 10-metastasis preliminary planning study with standardized VMAT SRS templates, extreme-M cluster arrangements produced measurable differences in normal brain and OAR doses, supporting M as a quick clustering tool. Ongoing work includes broader validation across additional cases and number of clusters and refinement of the metric formulation.