Predicting position of a functional target from an external marker in radiotherapy

The main goal of radiotherapy is to destroy the tumor while minimizing harm to nearby healthy tissue. Advances in the digital control have enabled planning and performing accurate treatments. However, todays technology is unable to compensate respiration induced motion, and therefore, ensure sufficient precision. One of the tasks in compensating respiratory motion is predicting position of the functional target (tumor) from an external marker during fraction. Performance of techniques, such as Pearson correlation, Gaussian filters, Fourries transformation, cross correlation, linear interpolation and partial-least squares, still leave plenty space for the improvement. We reports results of work in progress, i.e. experiments of applying different types of regressions to predict motion of functional target from different external markers. Results seem to be promising in most of the cases.