Room: Exhibit Hall | Forum 7
Purpose: Proton therapy is sensitive to anatomical changes and setup variation in the beam path. The purpose of this study is to find the adequacy of the CBCT in predicting changes in the dose distribution, hence reducing the times for QACT.
Methods: A cohort of 7 head neck patients, consisting of 13 CBCT scans, as well as the corresponding â€˜same-dayâ€™ QACT scans (as reference) were used. Both the TPCT and QACT scans were registered to the CBCT to represent the patientâ€™s position and anatomy during the treatment, using Varianâ€™s Velocity DIR algorithm. Some of the deformed TPCT images were further corrected using density override for internal tumor shrinks, cavity fillings. The CBCT images were also corrected using a deep-learning algorithm so that the image quality and HU accuracy are suitable for direct proton dose calculation. The mean absolute error (MAE) of HU was calculated between the cCBCT and QACT images. The initial planning beam set was applied to all three of these new image sets. The dose distributions were compared in terms of CTV V100, D95, and D1.
Results: The MAE is 18.11Â±4.45 HU between the cCBCT and QACT. The differences between the corrected DIR-based TPCT and QACT for the V100, D95, and D1 are 0.08%Â±0.75%, -0.01%Â±0.37%, and -0.34%Â±1.02%. The corresponding differences between cCBCT and QACT are 0.06%Â±2.00%, 0.00%Â±0.80%, and 0.07%Â±0.70%.
Conclusion: The dose calculation based on DIR TPCT can accurately reflect patientsâ€™ external and internal anatomy changes after manual density override method. The deep-learning based cCBCT can provide accurate dosimetric information. Both CBCT-based methods can be used to reduce QACT frequencies for patients treated with proton therapy.