A Framework For Inverse Treatment Planning On Perfexion Using Randomly Generated Isocentre LocationsKeywords: radiosurgery, technique, gamma knife, dose planning, physicsInteractive Manuscript
Ask Questions of this Manuscript:
What is the background behind your study?
Treatment planning for Perfexion can involve a challengingly large number of variables. As the treatment volume increases, and with the possibility of having fractionated treatments and incorporating different planning objectives such as dose homogeneity, forward planning approaches may not be capable of yielding the optimal treatment plan.
What is the purpose of your study?
The purpose of this work was to develop a framework for inverse treatment planning based on optimization.
Describe your patient group.
Describe what you did.
In the first stage of the optimization, isocenter locations are randomly generated within the treated volume. In the second stage, the optimal collimator size and weighting of each sector of each isocenter is determined based on a quadratic cost function minimization. The initial cost function penalizes over-dosing to normal structures and both over-dosing and under-dosing to target structures, but can be adapted to incorporate dose-homogeneity or dose-volume based constraints. The cost function is minimized using two approaches: (1) a projected gradient method, which terminates when the relative improvement of the cost function in successive iterations drops below a user-defined tolerance; and (2) semi-infinite linear programming (SILP) techniques. This model is tested on sample cases using pre-calculated dose grid kernels generated in a treatment planning system simulator. The test cases include gross tumor volume (GTV), planning target volume (PTV), and critical structures.
Describe your main findings.
The optimization algorithm achieved the objective of the 100% of GTV receiving at least the prescription dose, 95% of the PTV receiving at least 95% of the prescription, and critical structures receiving minimal doses. The SILP method outperformed the projected gradient algorithm in terms of computation time, and the resulting treatments were comparable to or better than the projected gradient solutions. Increasing the number of isocenters improves the resulting distribution though the rate of improvement decreases with the number of isocenters. Other optimization algorithms that may be more efficient and can incorporate data uncertainty are being explored.
Describe the main limitation of this study.
This is a retrospective study.
Describe your main conclusion.
An efficient and adaptable mathematical model for inverse planning for Perfexion has been developed.
Describe the importance of your findings and how they can be used by others.
Further improvements are expected to make the model more robust to data uncertainty, to improve the initial selection of isocenters, and to incorporate dose-volume and dose-homogeneity based constraints.
Treatment planning for Perfexion can involve a challengingly large number of variables. As the treatment volume increases, and with the possibility of having fractionated treatments and incorporating different planning objectives such as dose homogeneity, forward planning approaches may not be capable of yielding the optimal treatment plan.
The purpose of this work was to develop a framework for inverse treatment planning based on optimization.
In the first stage of the optimization, isocenter locations are randomly generated within the treated volume. In the second stage, the optimal collimator size and weighting of each sector of each isocenter is determined based on a quadratic cost function minimization. The initial cost function penalizes over-dosing to normal structures and both over-dosing and under-dosing to target structures, but can be adapted to incorporate dose-homogeneity or dose-volume based constraints. The cost function is minimized using two approaches: (1) a projected gradient method, which terminates when the relative improvement of the cost function in successive iterations drops below a user-defined tolerance; and (2) semi-infinite linear programming (SILP) techniques. This model is tested on sample cases using pre-calculated dose grid kernels generated in a treatment planning system simulator. The test cases include gross tumor volume (GTV), planning target volume (PTV), and critical structures.
The optimization algorithm achieved the objective of the 100% of GTV receiving at least the prescription dose, 95% of the PTV receiving at least 95% of the prescription, and critical structures receiving minimal doses. The SILP method outperformed the projected gradient algorithm in terms of computation time, and the resulting treatments were comparable to or better than the projected gradient solutions. Increasing the number of isocenters improves the resulting distribution though the rate of improvement decreases with the number of isocenters. Other optimization algorithms that may be more efficient and can incorporate data uncertainty are being explored.
This is a retrospective study.
An efficient and adaptable mathematical model for inverse planning for Perfexion has been developed.
Further improvements are expected to make the model more robust to data uncertainty, to improve the initial selection of isocenters, and to incorporate dose-volume and dose-homogeneity based constraints.
Project Roles:
