Optimal radiation beam profiles considering uncertainties in beam patient alignment.

The often large uncertainties that exist in beam patient alignment during radiation therapy may require modification of the incident beams to ensure an optimal delivered dose distribution to the target volume. This problem becomes increasingly severe when the required dose distribution of the incident beams becomes more heterogeneous. A simple analytical formula is derived for the case when the fraction number is high, and the desired relative dose variations are small. This formula adjusts the fluence distribution of the incident beam so that the resultant dose distribution will be as close as possible to the desired one considering the uncertainties in beam patient alignment. When sharp dose gradients are important, for instance at the border of the target volume, the problem is much more difficult. It is shown here that, if the tumor is surrounded by organs at risk, it is generally best to open up the field by about one standard deviation of the positional uncertainty--that is sigma/2 on each side of the target volume. In principle it is simultaneously desirable to increase the prescribed dose by a few per cent compared to the case where the positional uncertainty is negligible, in order to compensate for the rounded shoulders of the delivered dose distribution. When the tissues surrounding the tumor no longer are dose limiting even larger increases in field size may be advantageous. For more critical clinical situations the positional uncertainty may even limit the success of radiotherapy. In such cases one generally wants to create a steeper dose distribution than the underlying random Gaussian displacement process allows. The problem is then best handled by quantifying the treatment outcome under the influence of the stochastic process of patient misalignment. Either the coincidence with the desired dose distribution, or the expectation value of the probability of achieving complication-free tumor control is maximized under the influence of this stochastic process. It is shown that the most advantageous treatment is to apply beams that are either considerably widened or slightly widened and over flattened near the field edges for small and large fraction numbers respectively.