Abstract Book

S278

ESTRO 37

Purpose or Objective Day-to-day variations in respiratory motion caus e errors and uncertainties for external beam radiothera py. Breath-coaching methods using audio-visual feedback have been developed to try and make the respiration less variable, but these can be time-consuming to set up and require special equipment. This work investigates whether simple vocal instructions can be used to reduce the daily variability of the respiratory motion. Material and Methods The external respiratory motion was captured using a low-cost depth camera (Microsoft Kinect). Two respiratory signals were extracted from the 3D surface data, representing the displacement of the abdomen and thorax. 2D Gaussian Kernel Density Estimation (KDE) was used to model the relationship between the abdominal and thoracic signals during each session. The internal respiratory motion is thought to depend on both the abdominal and thoracic signals, therefore daily variations in the ‘abdominal-thoracic relationship’ will correspond to variations in the internal motion. The Correlation Coefficient (CC) was calculated between the KDE results from different sessions to assess the similarity in the ‘abdominal-thoracic relationship’ between those sessions, with a value of 1 indicating the relationship is exactly the same.Data was acquired on 5 days from 6 volunteers. On each day, data was first acquired for 3 minutes during Free-Breathing (FB), and then for 1 minute during Vocal instruction (V), where the subject was instructed to breath-in and breath-out with a fixed breathing rate of 4 seconds per breath. Results Fig.1 shows the KDE results for subject 1. It can be seen that in some FB sessions (top row) there are two parallel bands (indicating hysteresis between the two signals) but not in others. In the V sessions, there is only a single band, and the KDE results from the different sessions are more similar to each other.Table 1 gives the average CC values for each subject, when comparing every FB with every other FB, and likewise for the V. The vocal instruction made the KDE results more similar for subjects 1-3, but less similar for subject 4-6. However, analysing the results with a paired t-test showed that the results were only statistically significant (p-value < 0.05) for subjects 4 and 6, indicating the CC values could be very different

Table 1. The 2D CC for two sessions and the p-value of a paired t-test conducted to compare FB and V . OC-0526 Can surface guided radiation therapy be used to setup DIBH breast cancer patients to reduce Imaging? A. Moreira 1 , M. Zamburlini 1 , F. Cavelaars 1 , M. Guckenberger 1 , C. Linsenmeier 1 , I. Pytko 1 1 University Hospital Zürich, Department of Radiation Oncology, Thalwil, Switzerland Purpose or Objective Decrease of imaging dose as well as treatment time of breast cancer patients is of high interest in modern radiation therapy. Recently, surface guided radiation therapy (SGRT) to setup the patient based on their external body surface has become commercially available. This study aimed at testing if the setup accuracy of an SGRT system for the treatment of left sided breast cancer patients in DIBH is accurate enough to allow for a reduction of setup imaging for these patients. Material and Methods 17 left sided breast cancer patients with a total of 296 fractions were treated using a combination of 3D tangential and IMRT fields in DIBH. Patients were positioned supine on the breast board using an SGRT system (Align RT or OSMS). The patients’ reference surfaces (free breathing and DIBH) were imported from the CT scan and a region of interest of the treated area was selected. The patients were positioned using first the free breathing CT surface to establish the vertical baseline and then with the DIBH CT surface until all pre- shift deltas displayed on the SGRT systems were as close to zero as possible. After positioning, MV or KV imaging was performed in DIBH to verify the patient’s position. Using these images, an online bone match on the chest wall was performed and the patient position was corrected accordingly. All shifted values were recorded. Vertical (VRT), Lateral (LAT) and Longitudinal (LNG) shift values within 5mm were considered within our tolerance. This data was analyzed to conclude if patient setup with SGRT is accurate enough to reduce daily Imaging. Results Out of the 296 fractions, differences between SGRT and IGRT were within our tolerance of 5mm in all three directions in 277 fractions (93.6%), and within 6mm in 289 fractions (97.6%). The largest difference was observed in lateral direction; however, the most frequent axis that was out of tolerance was longitudinal (13 out of 19 failed fractions). This could be due to post set-up relaxation on the inclined breast board. Mean post-match shift based on imaging was -0.86 ± 1.55mm, 0.33 ± 2.40mm, and 0.65 ± 1.85mm in the VRT, LNG and LAT directions respectively. Conclusion The data shows that SGRT is accurate within 5mm in almost 94% of fractions for the setup of DIBH breast cancer patients. The fractions in which the shifts fell above tolerance occurred randomly, indicating lack of systematic error as well as lack of influence of breast swelling on the accuracy of SGRT in DIBH breast patients. In conclusion, the combination of daily patient positioning using SGRT and setup imaging performed only once a week would both reduce excess dose to the patient and treatment time on the Linear accelerator.

Fig. 1. 5 days of KDE for subject 1. X- and Y-axis represent the displacement of the abdomen and thorax. The a row represents FB sessions and the b row represents V sessions. Conclusion This work has presented a method for quantitatively assessing the daily variation in the relationship between abdominal and thoracic breathing traces, which is thought to correspond to variation in the internal respiratory motion. This has then been used to show that simple vocal instructions cannot reliably reduce the daily variation in the ‘abdominal thoracic relationship’, and in some cases, increases the variation.