Premium
MO‐E‐BRA‐03: Application of Robust Optimization in Lung Cancer Treatment
Author(s) -
Chan TCY,
Trofimov A,
Vrancic C,
Tsitsiklis JN,
Bortfeld T
Publication year - 2007
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.2761296
Subject(s) - robust optimization , probability density function , computer science , mathematical optimization , convolution (computer science) , motion (physics) , probability distribution , optimization problem , radiation treatment planning , mathematics , artificial intelligence , radiation therapy , statistics , medicine , artificial neural network
For lung tumors, the presence of motion due to breathing is a key source of uncertainty. Motion essentially blurs the static dose distribution, which can be thought of mathematically as a convolution of the static dose distribution with a probability density function (PDF) describing the motion. The 4D IMRT optimization/inverse planning method tries to undo this blurring effect by taking the motion PDF into account during the optimization of the intensity map. Such an approach is effective as a motion compensation technique, but only when the motion is highly reproducible over the entire treatment course. In terms of the PDF, “reproducibility” corresponds to witnessing the same PDF over the course of treatment that was observed in the treatment planning stage. However, if the realized PDF during treatment differs from the planning PDF, the subsequent convolution of the static dose distribution with the realized PDF may produce undesirable hot and cold spots. Robust optimization is a concept that has gained prominence in the optimization community for its wide applicability to problems with uncertain data. Real‐world problems are rarely, if ever, accompanied by noiseless data, hence, there is a natural motivation to incorporate this uncertainty into any optimization process. The robust framework we present builds on the 4D approach by explicitly accounting for the uncertain motion represented by uncertainty in the motion PDF. Instead of basing the optimization on one PDF, robust optimization uses a family of PDFs to create a static dose distribution that is less sensitive to variations in the motion. The robust framework allows us to craft solutions in the entire spectrum between the idealized 4D method, and a conservative, ITV‐like margin approach. A given intensity map that results from the robust optimization method will balance intensity‐modulation with intensity‐homogeneity in order to effectively trade off the sparing of healthy tissues with ensuring sufficient tumor coverage. Accordingly, the robust optimization method implicitly performs multi‐objective optimization on these competing objectives. Educational Objectives: 1. Understand the concept of robust optimization. 2. Understand the construction of robust treatment plans based on breathing motion PDFs. 3. Understand the mathematical and dosimetric differences between treatment plans of varying levels of robustness. 4. Understand the multi‐objective viewpoint of robust optimization.