PET Scanning Protocols for In-Situ Dose Delivery Verification of Proton Therapy

Positron emission tomography is so far the only method for in-vivo dose delivery verification in hadron therapy that is in clinical use. A PET scanner placed in the treatment position (in-situ) will be able to obtain the highest number of counts, as it minimizes the decay of the positron emitting nuclei before the scan is started as well as reduces the effect of biological washout. We investigated the influence of the scan protocol, i.e. the moment when a scan in done in relation to the treatment delivery, on the ability to measure unacceptable deviations from the treatment plan. We developed a Geant4-based Monte-Carlo framework for proton therapy simulations. Four patient cases are studied: two head-and-neck, one sarcoma near the spine, and one breast cancer case. For each irradiation field, the production of the following PET isotopes is calculated: 15O, 11C, 10C, 14O, 30P, 38K, and 13N. The time sequence of the pencil beam scanning irradiation, the decay of the PET isotopes during the irradiation, and biological washout are included in the simulation. The production of these nuclei is then used to calculate a PET image for two scan protocols: a scan of 120 seconds after the first field, or a scan of 120 seconds after the last field. To mimic a typical scanner spatial resolution, the images are blurred using a Gaussian blurring function with 4 mm FWHM. Deviations from the treatment plan are simulated by shifting the patient 4 mm perpendicular to the field angle, or by increasing the patient density by 3%. The PET image is then simulated again for each scanning protocol. The ability of each protocol to detect these deviations from the treatment plan is investigated by comparing the planned and the modified PET image. Several analysis methods are used: line profiles coupled to the field-directions, structural similarity index analysis [1], and gamma index analysis. [2] Preliminary data shows that a difference in density is best detected by starting the scan directly after the first field. However, shifts perpendicular to the field directions are better detected when the scan is done after the last field, due to the increased activity and counting-rate