Optimizing Dosimetric Indices for HDR Brachytherapy Using CVaR
Å Holm1*, T Larsson1, Å Carlsson Tedgren2, (1) Mathematical department, Linköping University, Linköping, Sweden, (2) Department of Medical Health Sciences, Radiation Physics, Linköping University, Linköping, SwedenSU-E-T-583 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: To develop a new model for automated optimization of HDR brachytherapy (BT) dose distributions that operates on the dose volume indices used in plan evaluation. The model can be solved to optimality, with constraints that are easy to interpret and modify for the clinical user. Recent research has shown that the optimization model hitherto used corresponds only weakly to such dosimetric indices. Alternative models that include such dosimetric indices have been presented; however, the explicit inclusion of dosimetric indices yields intractable models.
Methods: An approach for optimizing HDR BT dose distributions is presented where dosimetric indices are taken into account through surrogates based on the conditional value-at-risk concept (CVaR). This yields a linear optimization model that is easy to solve, and has the additional advantage that the constraints are easy to interpret and modify to obtain satisfactory dose distributions.
Results: Experimental comparisons have shown that our proposed model corresponds well with constraining dosimetric indices. All modifications of the parameters in our model yield the expected result. Both our approach and those presented earlier for optimization of dosimetric indices include an approximation, earlier models solve the models approximately, and we use a model based on an approximation. This gives us better control of the optimization. The dose distributions generated by the proposed optimization model are comparable to those generated by the standard model with respect to dosimetric indices.
Conclusion: Our new model is a viable surrogate to optimizing BT dosimetric indices that quickly and easily yield high-quality dose distributions. New optimization models that are capable of optimizing dosimetric indices will become of increasing importance in new applications involving more degrees of freedom than conventional HDR ¹⁹²Ir BT such as electronic BT.