The use of robust planning for proton radiotherapy to account for more complex changes than simple geometrical errors is at an explorative stage. In this project, we propose to expand the concept of automated robust planning to include robustness towards anatomical variations.

We propose to implement an individualized and progressive approach by characterizing the most prominent components of anatomical variation for the specific patient. Individual characteristics can be identified based on the daily imaging (cone-beam CT scans) performed in the treatment room for patient setup.

• Analysis and categorization of anatomical variations.

Method: Cone-beam CT scans performed on days of treatment (for setup purposes) will be used to quantify the variations for each patient. Deformable image registration will be performed between cone-beam CT scans and the treatment planning CT scan, resulting in a deformable vector field map for each registration. Daily anatomical variations will be quantified for each patient using a principal component analysis. Geometrical and physiological interpretation of the modes of variation will be performed by visual inspection and cluster analysis of the scans and the corresponding deformation fields.

• Individualization of adaptation and robustness strategy

The modes of anatomical variation and their interpretation are combined with recalculated dose distributions, to identify which modes of variation are most prone to result in large dosimetric changes. Based on these variations different approaches to individualized progressive robustness optimization and adaptation are compared. Strategies of increasing complexity range from inclusion of average variations at selected stages of treatment, to full progressively optimized plans with individual robustness measures based on daily variations.