In-beam pet monitoring for adaptive proton therapy: A voxel-based morphometry approach based on FLUKA Monte Carlo simulations.

One of the modalities for $in vivo$, non-invasive treatment monitoring in proton therapy is in-beam PET. Yet, no straightforward method exists to translate the information from PET images into easy to interpret information for physicians. We propose a new approach for analyzing in-beam PET monitoring images to locate, quantify and visualize morphological changes occurring over the treatment course. We thus selected a patient treated with proton therapy whose sinonasal cavity emptied over the treatment course. We generated a series of artificially modified CT scans to mimic this gradual emptying. Using a dual-head PET system geometry, we performed Monte Carlo simulations of the treatment to investigate monitoring-PET signal changes. Next, we performed two-tailed voxel-wise statistical tests to locate and quantify regions with significant morphological changes. As a result, we successfully applied our method to simulated in-beam PET images, generating three-dimensional probability maps to visualize morphological changes in the CT. Their characteristic color patterns are valuable to trigger an alarm in case of morphological changes over the treatment course.