Implementation of a fast algorithm for 3-D
inverse treatment planning for IMRT
Y. Xiao*
Thomas Jefferson University Hospital
Philadelphia, Pennsylvania
Deriving accurately and efficiently the optimized beam angle and beam
intensity pattern, given certain constraints for the tumor to be treated and the critical
organs to be spared, is the key to successful delivery of IMRT (Intensity Modulated
Radiation Therapy). The search for the optimized parameters consists of two processes:
dose calculation and the adjustment of parameters to achieve the objective. In this study,
we have used evenly spaced beam angles. For each beam angle, the beam opening is divided
into two-dimensional matrices of beam intensities. Dose calculation is performed for each
of these voxels using beam tracing and FFT convolution of electron transport kernels and
scatter kernels. A reduced size of the matrices is chosen in order to increase the
calculation speed while maintaining the calculation accuracy. Dose to the tumor and
critical organ due to each beam voxel is scored. The optimized weighting of beam voxels is
obtained by using the Cimmino simultaneous projection method, which solves a system of
inequalities without resorting to repetitious dose calculations. The time required for the
optimization of beam intensity of five 10x10 cm2 fields (1x1 cm2
voxels) with 0.5x0.5x0.5 cm3 dose resolution for the phantom is on the order of
5 minutes on a standard 300 Mhz pentium PC.
As an illustration of the calculation process, a common case is
selected which involves a 3-D prostate patient phantom with rectum as the critical organ
to be avoided. Five evenly distributed beam angles are pre-selected. For each angle, the
beam opening is divided into 10x10 cm2 intensity voxels.
Dose in the phantom due to each individual beam
intensity voxel is calculated using Fast Fourier Transform (FFT) with the truncated
kernel and the truncated phantom with the corresponding reduced matrix size
and with the beam voxel of interest as the center. Kernel tilting correction is easily
applied. Calculation time is greatly reduced. Dose for the voxels in the phantom due to
each beam voxel is stored in one matirx.
Dose limits are prescribed for the tumor volume and the critical organ to be
avoided. Namely, a set of inequality equations is established. The set of inequality
equations are solved using Cimmino algorithm:
The desired beam intensity map is thus obtained for certain beam angles
Abstracts Submitted to 1999 Annual AAPM Meeting | Medical Physics Home |
