K-means rapidly optimized isocenters and margins for polymetastatic stereotactic ablative radiotherapy

K-means matched hybrid optimization for margin volume while reducing median computation time from 306.9 seconds to 0.19 seconds.

KEY POINTS

  • This retrospective technical study evaluated 20 clinical thoracic cases containing 2–21 targets per patient, with maximum target separations of 3.8–32.4 cm.
  • K-means clustering generated candidate configurations ranging from one isocenter to one per target. Target-specific margins were calculated to maintain 95% coverage probability, assuming a baseline 5 mm translational margin and rotational uncertainties of 1°, 2°, and 3°.
  • Total margin volumes were closely aligned with the reference hybrid optimization method. Median differences were 0.03%, 0.14%, and 0.19% for rotational uncertainties of 1°, 2°, and 3°, respectively, with no significant differences.
  • Adding a second isocenter reduced median total margin volume by 31.0%, while moving from two to three isocenters produced a further 15.7% reduction. Benefits diminished with subsequent isocenters: 6.6%, 5.4%, and 3.7% for the fourth, fifth, and sixth isocenters.
  • Median computation time was 0.19 seconds with k-means versus 306.9 seconds for hybrid optimization. The study did not evaluate resulting dose distributions, treatment delivery, organ-at-risk exposure, or clinical outcomes.

CLINICAL TAKEAWAY

K-means clustering could allow planners to rapidly assess the trade-off between treatment efficiency, number of isocenters, and additional margin volume in complex multi-target thoracic stereotactic ablative radiotherapy. However, the framework was tested only in thoracic cases using modeled setup uncertainties and requires integration into treatment planning and dosimetric validation before clinical adoption.

SOURCE

Journal of Applied Clinical Medical Physics