(Sunday, 7/14/2019) 2:00 PM - 3:00 PM
Room: Stars at Night Ballroom 4
Purpose: Plans with multiple isocenters increase the complexity of the treatment and often use more MUs to deliver. However, they can improve target-dose conformity, especially for non-spherical/irregular/large tumors. We propose a mathematical model for efficient multiple-isocenter plans by optimizing the number and location of isocenters.
Methods: We formulate it as an optimization problem of choosing a set of ellipsoids that covers the target and minimizes the excess volume. To first determine the number of isocenters, we divide the target into several ball-shaped parts and the number of isocenters should match the number of such parts. We use the derivative of the curves to determine the number of balls. This gives us a good heuristic solution. To balance plan complexity and dose conformity, we set the maximum number of isocenters to be 3. A nonlinear-mixed integer program is used to determine the optimal placement.
Results: 25 patient cases were tested. The 5 prostate are ball-shaped and the algorithm returns 1 isocenter as expected. The 10 Lung tumors are large and irregular, 8 use more than 1 isocenter. For the 10 intracranial cases, although the tumors are tiny, they are highly irregular, with 6 cases require more than 1 isocenter. Overall, the treatment plans for all 25 cases uniformly result in improved conformity (9-15%). Since the radiation distribution is tighter around the tumor, the radiation dose to the OARs are greatly reduced also.
Conclusion: Our analysis shows that improved conformity can be achieved in all clinical cases. When ≤ 3 isocenters is used, irregularly-shaped tumors not only receive better conformity, the radiation to OARs also reduces. Although the problem is computationally challenging, our analysis shows that it can improve the quality of treatment plans significantly. Clinical trial has to be conducted to gauge the actual gain in treatment outcome.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
