Toward Robust Optimization of Adaptive Radiation Therapy
Michelle Böck will give a preview of her thesis that will be defended on the 14th of June.
A link to an online version of the thesis is given below the abstract.
Time: Mon 2019-06-10 11.00 - 12.00
Participating: Michelle Böck
Adaptive radiation therapy is an evolving cancer treatment approach which relies on adapting the treatment plan in response to patient-specific interfractional geometric variations occurring during the fractionated treatment. If those variations are not addressed through adaptive replanning, the resulting treatment quality may be compromised.
The purpose of this thesis is to introduce a conceptual framework that combines a variety of robust optimization approaches with the concept of adaptive radiation therapy. Robust optimization approaches are useful in radiation therapy, since interfractional geometric variations are accounted for while optimizing the treatment plan. Thus, combining these two concepts in a framework for robust adaptive radiation therapy gives the opportunity to optimize adapted robust plans which account for the actual interfractional variations in the individual case. In this thesis, a variety of frameworks with increasing complexity is introduced and their ability to handle interfractional variations is evaluated.
In the first paper, a framework based on the concept of combining stochastic minimax optimization with adaptive replanning is introduced. Within this framework, three adaptive strategies are evaluated based on their ability to mitigate the impact of interfractional variations on the accumulated dose. In these strategies, treatment plans are adapted in response to the measured variations by (i) modifying the probability distribution that governs the variations accounted for in the optimization, (ii) varying the level of conservativeness of the robust optimization approach, and (iii) modifying safety-margins around the tumor.
In the second paper, robust optimization approaches of varying levels of conservativeness are combined with optimization variables of varying degrees of freedom which account for fractionation and the interfractional geometric variations. The mathematical analysis shows that the solution of a time-independent problem is as good as the solution by the corresponding time-dependent problem, under the condition of convexity and independently and identically distributed interfractional geometric variations.
In the third paper, the framework from the second paper is extended to (i) handle unaccounted interfractional geometric variations with Bayesian inference, (ii) address adaptation cost through varying the adaptation frequency, and (iii) address computational tractability of robust optimization approaches with an approximation algorithm.
To emphasize the mathematical properties of the introduced frameworks, their performance is evaluated on an idealized one-dimensional phantom geometry subjected to a series of rigid translations. In this idealized phantom geometry, the relation between a modified optimization parameter and a feature in the resulting dose profile can be identified in a straightforward manner. This contributes to a better understanding of the underlying mechanisms between robustness, the adaptive strategies and the optimized dose profiles. The findings of this thesis are intended to provide a mathematical foundation for further development of the framework for, and research on, robust optimization of adaptive radiation therapy toward a clinical setting.
Link to Diva: online version