The objective of this research is to develop algorithms and software for treatment planning in intensity modulated radiation therapy under assumption of tumor and healthy organs motion. The current approach to addressing tumor motion in radiation therapy is to treat it as a problem and not as a therapeutic opportunity. However, it is possible that during tumor and healthy organs motion the tumor is better exposed for treatment, allowing for the prescribed dose treatment of the tumor (target) while reducing the exposure of healthy organs to radiation. The approach is to treat tumor and healthy organs motion as an opportunity to improve the treatment outcome, rather than as an obstacle that needs to be overcome. Intellectual Merit: The leading intellectual merit of this proposal is to develop treatment planning and delivery algorithms for motion-optimized intensity modulated radiation therapy that exploit differential organ motion to provide a dose distribution that surpasses the static case. This work will show that the proposed motion-optimized IMRT treatment planning paradigm provides superior dose distributions when compared to current state-of-the art motion management protocols. Broader Impact: Successful completion of the project will mark a major step for clinical applications of intensity modulated radiation therapy and will help to improve the quality of life of many cancer patients. The results could be integrated within existing devices and could be used for training of students and practitioners. The visualization software for dose accumulation could be used to train medical students in radiation therapy treatment planning.
Intensity modulated radiation therapy (IMRT) requires tight coordination between computational systems and the physical devices that deliver the prescribed treatment plan. The target tumor and surrounding regions of interest (ROIs) are often non-static due to motion introduced by the involuntary actions of the patient such as breathing and heartbeat. The current approach to addressing tumor motion in radiation therapy is to treat it as a problem and not as a therapeutic opportunity. Existing treatment planning methods attempt to create dose distributions that are at best dosimetrically equivalent to the static case. However, it is possible that during tumor and healthy organs motion the tumor is better exposed for treatment, allowing for the prescribed dose treatment of the tumor (target) while reducing the exposure of healthy organs to radiation. This project addressed a treatment planning and delivery paradigm that takes advantage of the differential motion of tumor and healthy organs to deliver the prescribed tumor dose while reducing the critical (healthy) structure dose relative to the static case. The research utilizes geometric optimization algorithms and applies them in radiation therapy studies. Investigation branches into a few major directions that include optimization of gantry movement, kinetic data structures to handle the motion of target and healthy tissue, GPU acceleration, and the simulation of multiple treatment plans. The project had the following main outcomes: (1) The effective delivery of various radiation therapy procedures requires fast motion of the gantry around the patient lying on the treatment couch and often fast change of speed of gantry in its angular motion. However, the gantry of a linear accelerator is also characterized by large inertia and changes in gantry angular speed have to be limited due to mechanical restrictions. The PIs have developed a solution to the problem of optimal efficiency delivery of therapy to moving targets that takes into account limitations of gantry acceleration. (2) Moving regions and kinetic data structures: If we know the motion of the target and the surrounding tissues and approximate the shape of the tissues as simple polygons, we can achieve better delivery by determining the time at which the amount of area overlap between the target polygon and the remaining non-target polygons is minimal. The time of minimal area overlap corresponds to maximum exposure of the target. This time (or interval of time) can be determined by maintaining a set of structures that defines the area of intersection between the target polygon and non-target polygons. We can approximate the motion of the tissue as a polyline (the flight plan of the polygon representing the tissue) and keep track of the area of intersection as each polygon moves along its flight plan in time. We have defined, implemented, and tested kinetic data structures to handle two dimensional motion. (3) Simulation of treatment planning: the PIs have developed a treatment visualization system that provides physicians with the ability to view and compare multiple radiation treatment scenarios at various levels of detail. Users are able to choose from a quad-view, which displays a given treatment at multiple levels of granularity at once, or an enlarged single granularity level, which is in 3D and fully navigable. Our system also allows the user to load multiple treatments from a database of patient treatments, and run their simulations side-by-side to compare their effectiveness and accumulated dosages over time. (4) GPU acceleration: an updated version of the simulator and an updated version of the tumor tracking algorithm have been implemented using GPUs. Since a significant part of the computation could be executed in parallel, this approach resulted in significant speed up in producing the sought solutions. (5) The PIs have addressed a number of theoretical problems that resulted from the proposed work. These include results on bichromatic and kinetic bichromatic separability of points, such as finding or maintaining a smallest circle that encloses all red points and the smallest possible number of blue points, and computing the convex hull for imprecise data points. (6) The PIs disseminated the results of this work at relevant conferences and workshops, and through presentations and talks at peer institutions.