Research is proposed to study a convolution modeling technique based on the fast Fourier transform (FFT) to compute dose distributions for conformal therapy. The goal of conformal therapy is to achieve higher local tumor control rates by shaping the dose distribution around the tumor so that the tolerance dose is determined by the normal tissue stroma within the target volume rather than by sensitive tissue structures outside the target volume. The work proposed here is aimed specifically at using the FFT convolution and other efficient algorithms to calculate the optimum dose distribution for a target volume and to generate the treatment parameters for the computer controls needed to produce the distribution. The research plan approaches the problem from two directions. One approach calculates the accumulated dose distribution by using FFT's to convolve an optimized axis weight distribution with a full rotation kernel. The beam weights and compensation as a function of gantry angle are calculated by ray-tracing from the axis weight distribution. The second approach calculates an ensemble of fields at twenty or thirty gantry angles using the single field FFT convolution model. Data abstracted from these distributions are used to search for a set of beam weights using an optimization algorithm that accounts for constraints placed upon the normal tissue structures. Both approaches are made feasible by the speed with which the FFT operates. Calculations operating on three- dimensional arrays that would take billions of seconds with conventional convolution algorithms require hundreds of seconds with the FFT. The accuracy of the algorithms will be checked against measured data and the speed of the two approaches will be compared. Algorithms such as these are essential to the effective implementation of technical improvements on therapy machines, such as computer control of treatment parameters and multileaf collimators.
Showing the most recent 10 out of 45 publications
