Prediction of radiation-induced normal tissue complications in radiotherapy using functional image data

The aim of this study has been to explicitly include the functional heterogeneity of an organ as a factor that contributes to the probability of complication of normal tissues following radiotherapy. Situations for which the inclusion of this information can be advantageous to the design of treatment plans are then investigated. A Java program has been implemented for this purpose. This makes use of a voxelated model of a patient, which is based on registered anatomical and functional data in order to enable functional voxel weighting. Using this model, the functional dose-volume histogram (fDVH) and the functional normal tissue complication probability (fNTCP) are then introduced as extensions to the conventional dose-volume histogram (DVH) and normal tissue complication probability (NTCP). In the presence of functional heterogeneity, these tools are physically more meaningful for plan evaluation than the traditional indices, as they incorporate additional information and are anticipated to show a better correlation with outcome. New parameters m(f), n(f) and TD(50f) are required to replace the m, n and TD(50) parameters. A range of plausible values was investigated, awaiting fitting of these new parameters to patient outcomes where functional data have been measured. As an example, the model is applied to two lung datasets utilizing accurately registered computed tomography (CT) and single photon emission computed tomography (SPECT) perfusion scans. Assuming a linear perfusion-function relationship, the biological index mean perfusion weighted lung dose (MPWLD) has been extracted from integration over outlined regions of interest. In agreement with the MPWLD ranking, the fNTCP predictions reveal that incorporation of functional imaging in radiotherapy treatment planning is most beneficial for organs with a large volume effect and large focal areas of dysfunction. There is, however, no additional advantage in cases presenting with homogeneous function. Although presented for lung radiotherapy, this model is general. It can also be applied to positron emission tomography (PET)-CT or functional magnetic resonance imaging (fMRI)-CT registered data and extended to the functional description of tumour control probability.