Room: Exhibit Hall | Forum 7
Purpose: To study the sensitivity of 3D EPID transit dosimetry to patient related variations, and to determine the optimal alert values for a set of indicators.
Methods: in air EPID measurements were used, in combination with the transmission calculated from patient CT data, to calculate virtual EPID transit images and to reconstruct virtual patient 3D dose distributions. 61 treatments were analysed: 11 VMAT prostate, 9 VMAT head-and-neck, 8 VMAT brain, 8 VMAT bladder and 25 IMRT lung. Setup errors were introduced by transforming the original CT data with translations in all directions (5mm to 20mm) and rotations along the cranio-caudal and left-right axises (4Â° to 16Â°). For the 25 IMRT lung treatments, changes in lung volume were introduced by uniformly contracting the lung contour (3mm to 12mm). The reconstructed 3D dose distributions were compared to the planned dose distributions (in the original CT) by Î³(3%/3mm) and DVH analysis using six indicators: Î”DISOC, Î³-mean, near Î³-max, Î³-pass rate, Î”PTV-D50 and Î”PTV-D98. For each combination of error and indicator, a receiver operator characteristic (ROC) curve was constructed, and the area under the ROC curve (AUC) was used to determine the error detectability.
Results: 3D EPID transit dosimetry is able to detect translation errors of 10mm (AUC=0.89), rotation errors of 8Â° (AUC=0.85) and lung contour contractions of 6mm (AUC=0.88). The detectability is highest for near Î³-max and Î³-pass rate for translation and rotations, and for Î”PTV-D98 for lung contour changes.
Conclusion: The detectability of patient related variations with 3D EPID transit dosimetry depends on the indicator used. Î³ analysis show on average better error detectability than DVH-based analysis and/or isocentre dose differences. ROC-based error detectability studies allow us to estimate optimal alert values for the clasification of errors observed with in vivo EPID dosimetry.