Analytic characterization of linear accelerator radiosurgery dose distributions for fast optimization

Linear accelerator (linac) radiosurgery utilizes non-coplanar are therapy d elivered through circular collimators. Generally, spherically symmetric are sets are used, resulting in nominally spherical dose distributions. Variou s treatment planning parameters may be manipulated to provide dose conforma tion to irregular lesions. Iterative manipulation of these variables can be a difficult and time-consuming task, because (a) understanding the effect of these parameters is complicated and (b) three-dimensional (3D) dose calc ulations are computationally expensive. This manipulation can be simplified , however, because the prescription isodose surface for all single isocentr e distributions can be approximated by conic sections. In this study, the e ffects of treatment planning parameter manipulation on the dimensions of th e treatment isodose surface were determined empirically. These dimensions w ere then fitted to analytic functions, assuming that the dose distributions were characterized as conic sections. These analytic functions allowed rea l-time approximation of the 3D isodose surface. Iterative plan optimization , either manual or automated, is achieved more efficiently using this real time approximation of the dose matrix. Subsequent to iterative plan optimiz ation, the analytic function is related back to the appropriate plan parame ters, and the dose distribution is determined using conventional dosimetry calculations. This provides a pseudo-inverse approach to radiosurgery optim ization, based solely on geometric considerations.