Backscatter for EPID dosimetry Tuesday Seminar Wonmo Sung Contents • • • • Introduction Literature Review SNUH clinic implementation Conclusion Introduction Electronic Portal Imaging Devices (EPID) : Amorphous silicon photodiode 1) For patient positioning 2) For DOSIMETRY! -> PDIP (Portal Dose Image Prediction) 1) 2) However… non-uniform support arm backscatter ? + How about backscatter from bunker? Literature Review Arm Backscatter • Physical approach lead sheet btw arm & EPID (2010) → Varian BSS-EPID development (2012) • Software approach backscatter kernel (2010) → (2013) + Bunker backscatter (2012) All of above research were conducted in the same research group (PI: Peter B. Greer, University of Newcastle, Australia) Backscatter (Rowshanfarzad et al, MP, 2010) EPID+ARM Devices: aS500 & E-arm (same physical components as the aS1000) For 40x30 cm2 field (EPID + arm) – (EPID) = Max 6% EPID Physical Approach - lead sheet - BSS-EPID Physical Approach - lead sheet (Rowshanfarzad et al, PMB, 2010) Optimal lead sheet : 2 mm (Backscatter reduction vs. weight) For 30x30 cm2 field, at CAX 6 MV 18 MV Lead 2 mm 0.1% 0.2% w/o lead 2.3% 1.3% Physical Approach - lead sheet (Rowshanfarzad et al, PMB, 2010) Symmetry improvement of in-line profiles Contrast-to-noise ratio & resolution are decreased. But not significant reduction in image quality Physical Approach - lead sheet (Rowshanfarzad et al, PMB, 2010) The displacement of the EPID central pixel was measured during a 360◦ gantry rotation with and without lead which was 1 pixel different. Physical Approach – Varian BSS-EPID (King et al, MP, 2012) • • • Varian developed BackScatter-Shielded (BSS)-EPID. BSS-EPID incorporates a layer of lead shielding beneath the active area of EPID Compare BSS-EPID + arm vs. BSS-EPID • For >99% pixels, 0.5% change • At CAX, maximum 0.2% effect on the central axis Physical Approach – Varian BSS-EPID (King et al, MP, 2012) • Use BSS-EPID to reconstruct dose in water for IMRT fields • Compare with MapCheck 2 cm depth, 2%/2 mm gamma Black dots: failed Software Approach - backscatter kernel 1) field size independent 2) field size dependent 3) dose conversion - just correct measurement (Varian updated) Software approach – backscatter kernel (Rowshanfarzad et al, MP, 2010) Rowshanfarzad developed simple backscatter kernel using measurements & Monte Carlo. EPID+ARM EPID Deconvolution +Gaussian fit Software approach – backscatter kernel (Rowshanfarzad et al, MP, 2010) BEFORE AFTER Gamma evaluation (2%/2 mm) for 20 × 20 cm2 field size Software approach – backscatter kernel (Rowshanfarzad et al, MP, 2010) 2%/2 mm Gamma evaluation, H&N IMRT Per Field Avg. 2% improvement with backscatter kernel Software approach – backscatter kernel FS dependent kernel…(Berry et al, MP, 2010) Kernel for dose reconstruction…(Wang et al., MP, 2013) Anyway, better kernel development so far. So far, we talked about… Energy fluence-convolution PDIP (Peter B. Greer, MP 36 (2009) 547-555) ~ similar to CCC 1) energy fluence (open beam+MLC leakage+head scatter) from Pinnacle 2) several kernels TERMA However, we use… Varian simple convolution PDIP (Ann Van Esch, Green 71 (2004) 223-234) 1) No energy spcetra condisderation 2) Single kernel to model 1) radiation source distribution 2) head scatter on the incident energy fluence, 3) effect of EPID scatter, 4) optical glare on the dose deposition in the EPID, 5) MLC leaf end transmission,… Varian PDIP algorithm (Ann Van Esch) Software approach – Just correct measurement! (Vinall et al, BJR, 2010) Ann accepted three major drawbacks of her algorithm + 1D correction 1) 40x40cm2 diagonal profile does not describe 40 x 30 cm2 field edge. 2) No backscatter correction 3) No spectral consideration (for head scatter, MLC…) She suggested 2D correction method to measurement data Please note that this is correction to measurement! Not to prediction as Peter did. Therefore…this method still possess problem. Software approach – Just correct measurement! (Vinall et al, BJR, 2010) R-arm, 40 × 30 cm2 field, in-line profile, Predicted : Calculation Official : Measurement + 1D-correction 2D no BS : Measurement + 2D-correction + backscatter correction 2D with BS : Measurement + 2Dcorrection + no backscatter correction Software approach – Just correct measurement! (Vinall et al, BJR, 2010) R-arm, 10 × 10 cm2 field, in-line profile, Predicted : Calculation Official : Measurement + 1D-correction 2D no BS : Measurement + 2D-correction + backscatter correction 2D with BS : Measurement + 2Dcorrection + no backscatter correction Bunker backscatter Bunker Backscatter (Rowshanfarzad et al, JACMP, 2012) 1) Portable brick wall experiment 2) Response variation with various gantry angle (MatriXX: output variation) Distance btw isocenter – left wall:370, right wall:385, floor:130, ceiling 145 cm 3) “Worst case scenario” : SDD=150+phantom, relative differences within 1.0% 1) Bunker Backscatter (Rowshanfarzad et al, JACMP, 2012) 1) Portable brick wall experiment 2) Response variation with various gantry angle (MatriXX: output variation) Distance btw isocenter – left wall:370, right wall:385, floor:130, ceiling 145 cm 3) “Worst case scenario” : SDD=150+phantom, relative differences within 1.0% 2) Bunker Backscatter (Rowshanfarzad et al, JACMP, 2012) 1) Portable brick wall experiment 2) Response variation with various gantry angle (MatriXX: output variation) Distance btw isocenter – left wall:370, right wall:385, floor:130, ceiling 145 cm 3) “Worst case scenario” : SDD=150+phantom, relative differences within 1.0% 3) SNUH clinic implementation • 방사선치료임상실습 (805.611 Practicum in Clinical Radiation Therapy) Instructor : 예성준 • 10-11 Weeks : three clinic-related topics Lecturer : 김정인 ① OSLD sensitivity trends : 박소연, 박종인 ② MatriXX commissioning : 이재기, 이현석 ③ PDIP commissioning : 성원모, 허태민 Varian PDIP commissioning procedure EPID calibration Imager Dark & Flood field calibration Dosimetry calibration Profile Correction Dmax 1D diagonal profile (w2CAD) Beam data import Output Factor measurement Actual Fluence measurement Optimal & Actual fluence Intensity profile Copy from PBC model Varian Update (CTB-PV-887-A, Oct. 15th, 2012 released) (a) The pre-configured PDIP algorithm for import into ‘Beam Configuration’. (b) A file for 2D profile correction including backscatter correction of the support arm imported during ‘Dosimetry Calibration’ on the treatment console. EPID calibration Imager Dark & Flood field calibration Dosimetry calibration Profile Correction Beam data import (b) Dmax 2D profile including 1D diagonal profile (w2CAD) backscatter correction Output Factor measurement Actual Fluence measurement Optimal Actual fluence (a) Golden Beam& Data Intensity profile Copy from PBC model Varian Update (CTB-PV-887-A, Oct. 15th, 2012 released) Using Golden beam data means… Golden EPID output factor & Golden kernel <My opinion - limitation> 1) Machines specific characteristics Ex. Same EPID? Same Linac? Varian Update (CTB-PV-887-A, Oct. 15th, 2012 released) G 2D profile correction to measurement including backscatter provided by Varian <My opinion - limitation> 1) field-size dependency on backscatter 2) Imager at off-axis Possible choices Prediction Measurement Varian Previous Approach P1 1D-corrected Measured Beam Data P2 2D-corrected Measured Beam Data PG 1D-corrected 1D-corrected Measurement Measurement M1 2D-corrected Measurement M2 Golden Beam Varian Updated Approach Preliminary results Machine : Varian Trilogy 6X, aS1000, E-arm Varian provided plan Relative dose normalized at isocenter ROI selection : Threshold of 5% Gamma Pass rate (%, 3 mm / 3% criteria, Local) P1 - M1 P1 - M2 PG - M1 PG - M2 IMRT1 84.1 82.2 91.5 93.5 IMRT2 91.5 94.3 91.7 94.2 VMAT1 72.2 76 65.6 69.5 VMAT2 95.3 91.7 98.7 97.9 Average 85.775 86.05 86.875 88.775 So far, “Varian previous approach” was the worst and “Varian updated approach” was the best. However, NOT ALWAYS. Future plan Re-Commissioning 1D-corrected measured Beam Data Commissioning 2D-corrected measured Beam Data Select 10 H&N clinical cases Mapcheck measurements to verify Trilogy TPS parameters (?) EPID measurements for 1D-corrected & 2D-corrected 10 H&N Conclusions Further improvement in the agreement between the model prediction and EPID measurements is required. More investigation is necessary for SNUH clinic implementation. Thanks for your listening! Any questions?