A neural network model for linear programming was proposed to optimize radiotherapy treatment planning. Comparing with traditional methods, neural networks exhibit notable robustness since their functions are not affected by parameter variations over a wide range. Linear programming based on neural networks (LPNN) can speed convergence, decrease scope of problems, and cut down extra slack variables and surplus variables. An example of the use of LPNN in three-dimensional stereotactic radiotherapy treatment planning was also described.
