Monte Carlo-based treatment planning for boron neutron capture therapy using custom designed models automatically generated from CT data.

PURPOSE A Monte Carlo-based treatment planning code for boron neutron capture therapy (BNCT), called NCTPLAN, has been developed in support of the New England Medical Center-Massachusetts Institute of Technology program in BNCT. This code has been used to plan BNCT irradiations in an ongoing peripheral melanoma BNCT protocol. The concept and design of the code is described and illustrative applications are presented. METHODS AND MATERIALS NCTPLAN uses thin-slice Computed Tomography (CT) image data to automatically create a heterogeneous multimaterial model of the relevant body part, which is then used as input to a Monte Carlo simulation code, MCNP, to derive distributions within the model. Results are displayed as isocontours superimposed on precisely corresponding CT images of the body part. Currently the computational slowness of the dose calculations precludes efficient treatment planning per se, but does provide the radiation oncologist with a preview of the doses that will be delivered to tumors and to various normal tissues, and permits neutron irradiation times in Megawatt-minutes (MW-min) to be calculated for specific dose prescriptions. The validation of the NCTPLAN results by experimental mixed-field dosimetry is presented. A typical application involving a cranial parallel-opposed epithermal neutron beam irradiation of a human subject with a glioblastoma multiforme is illustrated showing relative biological effectiveness-isodose (RBE) distributions in normal CNS structures and in brain tumors. Parametric curves for the MITR-II M67 epithermal neutron beam, showing the gain factors (gain factor = minimum tumor dose/maximum normal brain dose) for various combinations of boron concentrations in tumor and in normal brain, are presented. RESULTS The NCTPLAN code provides good computational agreement with experimental measurements for all dose components along the neutron beam central axis in a head phantom. For the M67 epithermal beam the gain factor for 1, boronophenylalanine for a small midline brain tumor under typical distribution assumptions is 1.4-1.8 x . Implementation of the code under clinical conditions is demonstrated. CONCLUSION The NCTPLAN code has been shown to be well suited to treatment-planning applications in BNCT. Comparison of computationally derived dose distributions in a phantom compared with experimental measurements demonstrates good agreement. Automatic superposition of isodose contours with corresponding CT image data provides the ability to evaluate BNCT doses to tumor and to normal structures. Calculation of gain factors suggests that for the M67 epithermal neutron beam, more advantage is gained from increasing boron concentrations in tumor than from increasing the boron tumor-to-normal brain ratio.