A number of treatment sites for external-beam radiation therapy, such as lung, breast, pancreatic, and liver cancers, move as the patient breathes, which compromises the precision of their irradiation. Modern radiation treatment modalities attempt to deal with this by adapting the radiation delivery to the respiratory motion as it occurs. This requires system control processes that can detect and
... [Show full abstract] anticipate respiratory movement patterns on a patient-by-patient basis in real time. Because breathing can be very idiosyncratic, this problem is a good candidate for machine learning algorithms that can be trained to model individual breathing patterns. Neural networks have proven quite effective in this capacity. This chapter describes the nature of the motion-compensated treatment problem and the issues in using a neural network to handle it.