The primary objective of this project is the development of a fast an accurate radiotherapy treatment planning system that is capable of producing automatically optimized radiation treatment plans, in real-time, for radiotherapy patients. In order to obtain maximum kill of tumor cells, it is desirable to deliver as high as possible a radiation dose to the tumor, subject to the constraint that the tolerance levels of healthy organs to radiation doses are not exceeded. This can be achieved by optimally arranging the radiation beams, or portals, that are used to deliver the radiation dose, as well as varying the intensity profile of the radiation beam within each portal, commonly known as intensity modulated radiation therapy, so that the ratio of radiation doses delivered to the tumor and the healthy organs are maximized. Currently, radiotherapy treatment planning is a labor- and computationally-intensive task due to the fact that a human operator is necessary for producing an optimized treatment plan or arrangement of radiation beams through an iterative trial and error process, and that more accurate dose calculation algorithms require computational power beyond what is commonly available in a radiotherapy treatment clinic. This proposed research project will therefore aim to develop computational methods that maximize the computational speed of accurate radiotherapy (external beam as well as brachytherapy) dose calculation algorithms. These fast and accurate algorithms will allow us to develop a method for the automatic real-time optimization of radiation beam arrangements. The algorithms will be benchmarked against both standardized test data and experimentally measured data in both homogeneous and heterogeneous phantoms that are used for clinical validation in radiation therapy. We propose to adopt the hierarchical partitioning method in achieving fast external beam and brachytherapy dose calculations. Coupled with small scale parallelism as available from commercial multiprocessor workstations and/or workstation networks, this method promises to enable the development of the type of radiotherapy treatment planning system that is direly needed in a state-of-the-art clinic setting, as indicated by the results from the preliminary studies that we have performed. In order to ensure the accuracy of the dose calculation results, we propose to modify the 3-dimensional convolution dose calculation algorithm for external radiotherapy and Monte Carlo simulation for brachytherapy by hierarchical partitioning of the desired dose calculation volume into disjoint subvolumes so that these currently time- consuming but highly accurate algorithms can be used in real-time for the treatment planning of radiotherapy patients. Finally, the hierarchical portioning method will again be used to divide the domain of optimization for treatment planning into subregions in order to achieve real-time optimized treatment planning, utilizing the accurate and fast dose calculation algorithms that we aim to develop.
