Maximizing the probability of satisfying the clinical goals in radiation therapy treatment planning under setup uncertainty

• Published in: Medical physics (Lancaster) Vol. 42; no. 7; pp. 3992 - 3999
• Main Authors: Fredriksson, Albin, Forsgren, Anders, Hårdemark, Björn
• Format: Journal Article
• Language: English
• Published: United States American Association of Physicists in Medicine 01.07.2015
• Subjects: 07 ISOTOPES AND RADIATION SOURCES; biological organs; COMPARATIVE EVALUATIONS; Digital computing or data processing equipment or methods, specially adapted for specific applications; Dosimetry; error analysis; geometry; Humans; IMPT; IMRT; In which a programme is changed according to experience gained by the computer itself during a complete run; Learning machines; Inference methods or devices; Intensity modulated radiation therapy; Male; medical computing; Medical treatment planning; optimisation; OPTIMIZATION; PHOTON BEAMS; Photons; Photons - therapeutic use; PLANNING; PROBABILITY; Prognosis; PROSTATE; Prostate - radiation effects; PROTON BEAMS; Proton therapy; Proton Therapy - methods; Protons; Protons - therapeutic use; radiation therapy; Radiation treatment; Radiometry - methods; RADIOTHERAPY; Radiotherapy Planning, Computer-Assisted - methods; Radiotherapy, Intensity-Modulated - methods; robustness; set theory; Therapeutic applications, including brachytherapy; Treatment Outcome; Uncertainty; uncertainty handling
• ISSN: 0094-2405; 2473-4209; 2473-4209
• DOI: 10.1118/1.4921998

Purpose: This paper introduces a method that maximizes the probability of satisfying the clinical goals in intensity‐modulated radiation therapy treatments subject to setup uncertainty. Methods: The authors perform robust optimization in which the clinical goals are constrained to be satisfied whenever the setup error falls within an uncertainty set. The shape of the uncertainty set is included as a variable in the optimization. The goal of the optimization is to modify the shape of the uncertainty set in order to maximize the probability that the setup error will fall within the modified set. Because the constraints enforce the clinical goals to be satisfied under all setup errors within the uncertainty set, this is equivalent to maximizing the probability of satisfying the clinical goals. This type of robust optimization is studied with respect to photon and proton therapy applied to a prostate case and compared to robust optimization using an a priori defined uncertainty set. Results: Slight reductions of the uncertainty sets resulted in plans that satisfied a larger number of clinical goals than optimization with respect to a priori defined uncertainty sets, both within the reduced uncertainty sets and within the a priori, nonreduced, uncertainty sets. For the prostate case, the plans taking reduced uncertainty sets into account satisfied 1.4 (photons) and 1.5 (protons) times as many clinical goals over the scenarios as the method taking a priori uncertainty sets into account. Conclusions: Reducing the uncertainty sets enabled the optimization to find better solutions with respect to the errors within the reduced as well as the nonreduced uncertainty sets and thereby achieve higher probability of satisfying the clinical goals. This shows that asking for a little less in the optimization sometimes leads to better overall plan quality.