Implementations of the generalised Gaussian pencil beam algorithm for three-dimensional electron beam dose planning: Dissertation

The use of high energy electron beams in radiation therapy has achieved an important role during the recent years. New requirements have been set for computerized dose planning, both in accuracy, in three-dimensionality, and in speed. In this work the computation of absorbed dose distribution for electron beams in radiation therapy was studied. The aim was to improve the computation in the three-dimensional patient geometries with complex radiation beam settings. This work was started as part of the Nordic CART project for the development of the use of imformation technology in radiation therapy carried out during the years 1985-87. The development of the treatment units producing electron beams, the dose computation algorithms for electron beams, and the principles of the clinical use of electron beams are reviewed. The theory of the generalized Gaussian pencil beam algorithm was applied for computing the dose distributions for electron beams in the CT based dose planning system, SAITOM, with implementations for rotated collimator asymmetric collimation, irregular field shapes and separate blocks in the field and computation in off axis slices. The accuracy of the dose computation was evaluated by comparing the computed dose distributions to measurements in a water phantom made in the 14 MeV electron beam of a Microtron MM14 accelerator. An implementation of the algorithm was also made for an array processor (FPS 5110) to study the separation of the algorithm for parallel processing and to test the achieved speed for applications in three dimensional dose computation. The clinical applicability of the implementations was demonstrated with two typical clinical patient treatments. The accuracy of the implementations was generally within ± 2 mm and ± 3 % for rectangular fields and within ± 2 mm and ± 5 % for irregular fields with blocks. The algorithm was clearly separable for parallel processing between the host computer and the array processor with almost equal computing times for the two parts. The achieved increase in speed of computation with the AP in the test conditions was 11.5- fold. The implementations of the generalised Gaussian pencil beam algorithm for the new conditions are sufficiently accurate to allow the use of all options in the treatment units with reliable computations for blocked fields. The algorithm separates optimally for implementations for an array processor. The achieved speed of dose computation with the AP was acceptable for three dimensional computation of the dose distribultion. The developed implementations offer advantages for clinical radiation therapy treatment planning by improving the accuracy of the dose computation for irregular fields and under and close to blocks. The implementations do not, however, take into account the effects of inhomogeneities in three dimensions, which will be the next step in the development of the pencil beam algorithms.