An evolutionary algorithm for optimization of affine transformation parameters for dose matrix warping in patient-specific quality assurance of radiotherapy dose distributions

Patient-specific quality assurance for sophisticated treatment delivery modalities in radiation oncology, such as intensity-modulated radiation therapy (IMRT), necessitates the measurement of the delivered dose distribution, and the subsequent comparison of the measured dose distribution with that calculated by the treatment planning software. The degree to which the calculated dose distribution is reproduced upon delivery is an indication that the treatment received by the patient is acceptable. A new method for the comparison between the planned and delivered dose distributions is introduced; it assesses the agreement between planar two-dimensional dose distributions. The method uses an evolutionary algorithm to optimize the parameters of the affine transformation, consisting of translation, rotation, and dilation (or erosion), that will warp the measured dose distribution to the planned dose distribution. The deviation of the composite transformation matrix from the identity matrix is an indication of an average geometrical error in the delivered dose distribution, and represents a systematic error in dose delivery. The residual errors in radiation dose at specific points are local errors in dose delivery. The method was applied to IMRT fields consisting of horizontal intensity bands. Analysis shows the method to be a promising tool for patient-specific quality assurance.