ESTRO 38 Abstract book

S947 ESTRO 38

EP-1755 Implementation of EPID in vivo dosimetry for SBRT: setting tolerance levels for routine clinical use M. Esposito 1 , A. Ghirelli 1 , S. Pini 1 , S. Russo 1 , G. Zatelli 1 , P. Alpi 2 , R. Barca 2 , B. Grilli Leonulli 2 , L. Paoletti 1 , F. Rossi 1 , P. Bastiani 2 1 Azienda USL Toscana Centro, S.C. Fisica Sanitaria, Bagno a Ripoli, Italy ; 2 Azienda USL Toscana Centro, S.C. Radioterapia, Bagno a Ripoli, Italy Purpose or Objective Electronic portal imaging device (EPID) transit in vivo dosimetry is a powerful tool for patient specific quality assurance (QA), however there is a lack of consensus about which dose difference metric should be used for alert and action level. The aim of this work is to analyze results of transit EPID dosimetry of stereotactic treatment in the abdominal and pelvic sites, and establish tolerance levels for clinical routine use that are sensitive enough to find relevant errors but have also high specificity. Material and Methods 64 stereotactic VMAT treatments (113 fractions) with target in the abdomen or pelvis were analyzed (15 liver, 10 adrenal gland, 7 spine, 32 lymph nodes). In vivo 3d doses were reconstructed with the back-projection EPID commercial algorithm Dosimetry Check 4.10. The differences between planned and in vivo doses were evaluated using Gamma Agreement Index (GAI) 3%/3mm (20% Dose threshold), and dose volume histogram (DVH) differences in prescription target volume (PTV) and clinical target volume (CTV). Initial tolerance levels were set to GAI<85% in PTV. Fractions exceeding tolerance levels (OTL) were checked by experienced Medical Physicists and were classified as due to set-up errors, incorrect use of immobilization devices, 4d errors, transit EPID algorithmic error, and unknown/unidentified errors. Results 44% of total fractions were out of tolerance levels and were classified as: set-up errors (5%), incorrect immobilization device (3%), 4d errors (3%). 50% of OTL were due to transit EPID algorithm failure. In the remaining 28% the OTL causes were not identified. Average difference of PTV and CTV mean dose (± standard deviation) were -3.4%±3.4% and -2.4%±3.2% Average GAI (81.8±20.6) % in PTV and (72.8±31.2) % in CTV. Setting the tolerance level to ΔCTV mean dose > 5.5% the percentage of OTL decrease to 15% and only in one case (6%) EPID algorithmic error occurred.

by the Eclipse™ TPS in small static fields in homogeneous medium. Material and Methods We considered square fields in the range from 5 mm to 30 mm side produced by a multi-leaf collimator (MLC). Fields were centred on machine central axis (CAX) or off-CAX by 100 mm; collimators set to 0º or 90º. All fields were produced by a 6 MV flattening filter free beam delivered by a Varian TrueBeam STx® linear accelerator equipped with an HD-MLC. TPS plans were generated by Eclipse 15.5, independently commissioned and optimized for IMRT/VMAT deliveries. Calculations were performed, using the same beam model as input, a single isocentre and 1.0 mm grid, by the analytic anisotropic algorithm (AAA) and the Acuros-XB algorithm (AXB). Calculations were compared in terms of output factors (OFs) and off- axis ratios (OARs) (FWHM and penumbral width at 20%- 80%) against measurements with a prototype 2D solid- state dosimeter ‘Octa’ and Gafchromic™ EBT3 films. Results In a 10 mm field on CAX, calculated and measured OFs (Table 1) were in agreement within 1.4% (AAA) and 1.5% (AXB) with collimators at 90º, and within 1.7% (AAA) and 1.8% (AXB) with collimators at 0º. In this same field off- CAX, agreement was within 5.1% (AAA) and 8.9% (AXB). In- line FWHMs (Fig. 1) were overestimated by calculations, with largest differences in 5 mm and 10 mm fields. In the latter field, overestimations were of comparable magnitude regardless of collimators rotation and position with respect to CAX. In-line penumbral widths (Fig. 1) were overestimated by calculations, with largest differences in off-CAX fields.

Table1 In this table, results of the application of three different alert criteria are shown.

Conclusion Calculation errors, which may be enhanced in small static fields, are expected to be smoothed out in dynamic multi- field clinical planning. In the present study, differences between calculations and measurements were affected by field size, position with respect to CAX and collimators rotation. Wider calculated FWHMs and penumbral widths were likely the result of a beam modelling aimed at ensuring modulated fields were accurately delivered, at the potential expense of inaccuracies in static fields. Our results emphasized the necessity of a thorough verification of a TPS in small static fields. The study had the limitation of not investigating the influence of parameters such as different back-up jaws settings, beam energy and depths in water.