ESTRO 2020 Abstract book

S935 ESTRO 2020

cancelled if radial prostate movement exceeded 3 mm from the CBCT match point (the origin). The dose reconstruction was implemented using the scripting interface of a research version of RayStation 9A treatment planning system (RaySearch Laboratories AB, Stockholm, Sweden) reading 3D-coordinates of a RayPilot time series representing prostate movement, and k -means clustering up to 30 clusters. In this retrospective study, the prostate was allowed to move more than 3 mm. Dose distribution was reconstructed for each cluster and mapped back to the original CT image. CT images were created for each cluster by translation of a rigid prostate (CTV). Femoral heads were chosen to remain rigid and fixed. Soft tissues, and to some extent also pelvic bones, were deformed and transferred in relation to the prostate movement. Results Since the treatment system did not record beam on times in relation to the movement data, worst case scenarios were chosen to study what could have happened without real-time monitoring during one fraction. The method allowed prostate movement induced D95 reductions to be analysed per fraction series. Assuming a stationary prostate for the other four fractions, D95 reductions for the five patients were 7.6, 19.4, 0.3, 0.1 and 4.4 %. This represents a minimum error for the treatment. A maximum error would occur if the prostate movement was similar in all the five fractions. In that case the D95 reductions would be 38, 97, 1, 1 and 22 %. Fig.1. Recorded prostate movement data. The crosshair represents the point of origin for movement data simulating a match based on CBCT image. The red and green stripes correspond to beam on times for the two arcs of the VMAT plan.

SGRT system (AlignRT, Vision RT) was used for initial patient setup and gating. A kV/MV image pair was acquired during breath-hold (BH) and matched with digitally reconstructed radiographs (DRRs) to determine the chest wall position of the day. The couch correction to account for daily variations was performed during a subsequent BH, and a new reference surface was acquired and used for the SGRT gating. This procedure was needed because a new SGRT reference surface cannot be captured simultaneously with the IGRT imaging. Continuous portal MV images (cine MV) recorded at 7.7Hz during all treatments were registered to DRRs post-treatment to determine the treatment error in beam’s eye view. The MV imaging and SGRT signals were synchronized and the following five error contributions were determined in BEV for each fraction: (1) on-line MV match error, (2) inter-breath-hold shift between setup imaging and couch correction, (3) intra-breath-hold patient motion from start of couch correction to capturing of the new reference surface, (4) difference between actual and intended SGRT DIBH level during treatment, and (5) intrafraction shift of the chest wall relative to the SGRT surface. In addition to our current setup protocol (Scenario A), two alternatives were simulated after correction for error (1) above (to remove operator-dependent errors in the comparison): (B) image- guided SGRT with a daily reference surface captured after imaging (errors 2-5), and (C) image-guided SGRT with the reference surface captured at the time of imaging (errors 4-5). Results The SGRT reported error in general correlated well with the actual error in cine MV images at a fraction, but fraction specific offsets were present due to errors introduced in the setup procedure (Figure 1). Table 1 summarizes the individual error contributions and scenario accuracies. Scenario C led to the best result. All scenarios were statistically different from each other (p<0.05). Conclusion A detailed analysis of the error budget in image-guided SGRT of breast cancer was performed. Two alternative SGRT scenarios were investigated. Image guidance can improve SGRT, but for optimal accuracy the reference surface of the day should be captured simultaneously with IGRT imaging. PO-1620 Reconstruction of intra-fractional real-time prostate movement and its effect on dose distribution L. Järvinen 1 , M. Tenhunen 1 , M. Myllykangas 1 , M. Persson 2 , E. Traneus 2 , O. Sjöberg 3 , P. Ekström 3 1 Helsinki University Hospital, Cancer Center, Helsinki, Finland ; 2 Raysearch Laboratories AB, Raysearch Laboratories AB, Stockholm, Sweden ; 3 Micropos Medical AB, Micropos Medical AB, Gothenburg, Sweden Purpose or Objective Stereotactic body radiation therapy (SBRT) of prostate cancer (PCa) has recently been shown well tolerated in phase I and II trials indicating favorable normal tissue complication probability. An ongoing discussion suggests that the α/β ratio of PCa could be even lower than the α/β ratios of the surrounding organs, which supports use of hypofractionation for increased tumor control probability (TCP). As the number of treatment fractions is reduced, the risk of a TCP compromising geometric miss increases, which is the focus of the present study. Material and Methods The selected five patients with local PCa received 7 Gy twice a week in 5 fractions. Real-time monitoring of prostate motion was manifested using the RayPilot system (Micropos Medical AB, Gothenburg, Sweden), which consists of a treatment table top overlay containing an integrated receiving antenna array, and a wired transperineally implanted electromagnetic transmitter. The target was matched using a CBCT. Treatment was

Fig.2. Original and reconstructed dose distributions of a cranial CT slice. The lack of dose in the PTV is the result of a cranial movement.