ESTRO 38 Abstract book

S573 ESTRO 38

The implications of this new method on SRS dosimetry are currently being explored. PO-1032 The potential of CBCT for setup and treatment verification in proton therapy for prostate cancer E. Fura 1 , A. Dasu 2 , A. Ureba 3 , U. Isacsson 4 , S. Johansson 5 1 Stockholm University, Medical radiation physics, Stockholm, Sweden ; 2 Skandionkliniken, Medical radiation physics, Uppsala, Sweden ; 3 Karolinska University Hospital, Medical Radiation Physics, stockholm, Sweden ; 4 Uppsala university hospital, Medical radiation physics, Uppsala, Sweden ; 5 Uppsala university hospital, Oncology, Uppsala, Sweden Purpose or Objective Proton beam treatments (PBT) in the pelvic region can prove difficult due to the sensitivity of the particle tracks to changes in the medium and its implications on the dose distribution. Various registration approaches and stabilisation techniques are available for setup verification for pelvic treatments such as rigid registration to structures like bones and markers, but studies on the matter are scarce. Cone Beam Computer Tomography (CBCT) provides 3D information on the morphology of the patient and could be used daily in the clinic as an image- guided system to verify the patient positioning. However, CBCT images cannot generally be used directly to investigate the impact of inter-fractional changes on the dose distribution. Therefore, this study aimed at developing a method for assessing the impact of inter- fractional changes in proton therapy for prostate cancer by means of CT-to-CBCT deformable registration for various registration approaches. Material and Methods Five prostate cancer patients with rectal rods, used as a stabilization technique, previously treated with photon therapy were included in this study. The patients were replanned with protons for a boost treatment aimed at delivering 4 fractions of 5 Gy, to a total of 20 Gy. The patients underwent a planning CT(pCT) before treatment and one CBCT before each fraction of the boost. The CBCTs were used to deform the pCT to four new deformed image sets representing each fraction. The newly deformed image sets were then rigidly registered to the pCT with two different registrations approaches, bones and markers. The robustness of these registrations was evaluated for two different plans on the pCT (a lateral field arrangement and an anterior oblique field arrangement). The plans were then copied and recalculated over to the bone or marker registered deformed CT sets. After deforming the dose distributions to the pCT, all fractions were then summed up, compared to the reference plan of the pCT and evaluated. Results Bone and marker registrations for both plans resulted in close to 100% coverage of the CTV. The doses to nearby OAR on the summed plans were below the constraints. The coverage of the PTVs on the summed plans were up to 8% lower than for the reference plan. Interestingly, there was a trend for the PTVs where the registration to the bones for the lateral field arrangement had consistently higher coverage than the registration to the markers for all patients. Conclusion The results show that it is possible to use deformable image registration of CBCTs to verify the coverage of the target with PBT. Registering the deformed CT to the bones of the pCT seems to be more robust than registration to the markers in the prostate when using a lateral field arrangement. However, both plans can be used clinically without compromising the coverage of the target.

Results A summary of the results for all fields tested can be seen in figure 2. The range of calculated minimum distances between the gantry and the patient or the treatment couch is shown for each couch angle. The average minimum distance was 16.6 cm and the median was 15.8 cm. One colliding field, using a 5 cm patient expansion zone, was detected by the software out of a total of 148 fields tested for 36 patients. This field was treatment simulated on a Linac and confirmed to be a collision risk. No undetected collisions occurred clinically during the implementation phase.

Conclusion We successfully clinically implemented collision detection software for SRS. The software was capable of detecting all potential collision risks. Moving forward, using the automated software instead of performing treatment simulation on a Linac will result in clinical resources being utilized more effectively and help avoid any potential re- planning due to collision risk. While the current conservative gantry trajectory treatment planning approach is effective at avoiding most collisions, a planning approach that utilizes patient specific model based collision testing will allow for a larger number of control points to be used in treatment plan optimization.