Boron Neutron Capture Therapy Treatment Planning Improvements

The Boron Neutron Capture Therapy (BNCT) treatment planning process of the Harvard/MIT team used for their clinical Phase I trials is very time consuming. If BNCT proves to be a successful treatment, this process must be made more efficient. Since the Monte Carlo treatment planning calculations were the most time consuming aspect of the treatment planning process, requiring more than thirty six hours for scoping calculations of three to five beams and final calculations for two beams, it was targeted for improvement. Three approaches were used to reduce the calculation times. A statistical uncertainty analysis was performed on doses rates and showed that a fewer number of particles could not be used and still meet uncertainty requirements in the region of interest. Unused features were removed and assumptions specific to the Harvard/MIT BNCT treatment planning calculations were hard wired into MCNP by Los Alamos personnel, resulting in a thirty percent decrease in runtimes. MCNP was also installed in parallel on the treatment planning computers, allowing a factor of improvement by roughly the number of computers linked together in parallel. After theses enhancements were made, the final executable, MCNPBNCT, was tested by comparing its calculated dose rates against the previously used executable, MCNPNEHD. Since the dose rates in close agreement, MCNPBNCT was adopted. The final runtime improvement to a single beam scoping run by linking the two 200MHz Pentium Pro computers was to reduce the wall clock runtime from 2 hours thirty minutes to fifty nine minutes. It is anticipated that the addition of ten 900 MHz CPUs will further reduce this calculation to three minutes, giving the medical physicist or radiation oncologist the freedom to use an iterative approach to try different radiation beam orientations to optimize treatment. Additional aspects of the treatment planning process were improved. The previously unrecognized phenomenon of peak dose movement during irradiation and its potential for overdosing the subject was identified. A method of predicting its occurrence was developed to prevent this from occurring. The calculated dose rate was also used to create dose volume histograms and volume averaged doses. These data suggest an alternative method for categorizing the subjects, rather than by peak tissue dose. Thesis Supervisor: Guido R. Solares Title: Assistant Professor of Radiology at Harvard Medical School Thesis Reader: Jacquelyn C. Yanch Title: Associate Professor of Nuclear Engineering and Whitaker College of Health Sciences and Technology Thesis Reader: Robert Zamenhof Title: Associate Professor of Radiology at Harvard Medical School