Image-based quantification of radiation-induced lung damage

The results presented in this thesis explored the associations of a quantitative lung damage endpoint (∆HU) with radiation dosimetry (sigmoidal dependence), clinical variables, regional location and a clinical dyspnea (change) endpoint. Furthermore, we showed the feasibility of a selective lung damage avoidance workflow based on the planning CT densities and made important recommendations to limit lung damage and dyspnea. Radiation dose to the heart was confirmed to be involved in lung damage induction, the volume effect (total lung V5) was associated both with dyspnea worsening and lung damage probability, and dose escalation in a randomized trial (the PET-boost trial) did not show unexpected increases in the objective lung damage metric. Interestingly, proton therapy comes with a potential advantage for all these dosimetric features. Refinement of our lung damage prediction model could come from advances in imaging (routine 4DCT acquisition), reproducible genetic markers of radiation toxicity and new knowledge of underlying biological phenomena (e.g. from small animal studies).