الفهرس 
COVEROnly 14 pages are availabe for public view
 |  | from | 185 |  |  |
| | | from | 185 | | |
Abstract
Electronic portal imaging devices (EPIDs) were initially developed for the purpose of patient setup verification. The use of EPID in dosimetry applications is increased day after day with the availability of amorphous silicon (aSi) EPIDs with their phosphor scintillation panel. The aim of the present work is to develop the use of an aSi-EPID in the dosimetric verification of intensity modulated radiation therapy (IMRT) technique. Special attention was given to the multileaf collimator (MLC) in the linear accelerator (Linac) head as it’s the main modulator and beam shaping device in the intensity modulation process.
Like any detector, QA tests had to be performed on the EPID. Pixel reading stability was studied over short and long periods. The standard deviation of detector readings, corrected for linac output variations, was 0.13%. Pre-irradiation of the detector was found to have a small influence on pixel reading (<1%), i.e. no warming up is required to start image acquisition. The pixel reading with no extra buildup was within 1% of the maximum reading, due to the inherent copper plate which acts as a build up layer. Image contrast resolution was optimized with different dose rates and found to be increased at higher dose rates.
The pixel reading was calibrated versus incident dose (with no phantom in-between), a linear calibration curve was established up to 300 MU within 1% accuracy at 6 and 15 MV. This curve will be used over this study for pixel reading dose response calculation. MATLAB environment was used to analyze portal images and estimate the 2D dose distribution at the isocenter plane. MATLAB was commissioned for portal dose calculation at 3×3 cm2 field size (to represents the IMRT conditions). The total number of MUs delivered was calculated by the treatment planning system (TPS) and transferred by Record and Verify (R/V) system to the linac console. If the measured dose distribution from portal images matches with TPS calculation this implies the correct transfer of all IMRT segments with the correct desired leaf sequences and doses which means safe treatment. To account for patient/phantom presence between the beam and the EPID, a set of phantom correction factors (pcf) for different phantom thicknesses, field sizes, and distances between phantom and EPID are used. These factors were all have a linear behaviour. Because of the thickness variation of the flattening filter, spectral variations occur across the field, the spectral variation increases as the field size decreases and as the offset increases. Most severest off axis offsets was found to be within 2.8% accurracy. Clinical demo reference IMRT prostate and breast planning were implemented to check the method. The 2D absolute dose distributions were calculated from portal images within 1% accuracy from TPS calculations. Furthermore, intensity modulation portal dose profiles matched with those obtained from the TPS.
Linac QA was also performed using portal images analysis by ImageJ NIH software, provided that the pixel element size is 0.25×0.25 mm2. The light and radiation field congruence was detected for a variety of field sizes and compared to films. Gantry Sag effect on the field size was found to be within 1 mm. The collimator and table rotation accuracy (within 2 mm) around the isocenter was detected at different angles and compared with films. In-air open field penumbra, effective penumbra, flatness and symmetry were tested and compared to ion chamber measurements, they all were within 1% accuracy. Wedge factors were calculated with 1% accuracy relative to the ion chamber measurements.
The introduction of MLC into clinical use added a new dimension to the complexity of the calibration and quality control checks. Primary collimator-MLC alignment was checked. Leaf position precision measurements were compared to films to accuracy of 0.5 mm. MLC Leakage, transmission and tongue_groove effects were measured and compared to film and ion chamber measurements within an accuracy of less than 1%. Using the MATLAB and portal images, we exploited the unique Varian® MLC Dynalog file to create an analogous file for Elekta® MLC. Reference MLC patterns were designed to commission the method, leaf-end positions were calculated as the locations where the image intensity is 50% of the maximum. Portal images extracted leaf positions were compared to those recorded in the corresponding Dynalog files. The results showed that the EPID-extracted leaf-end positions were within ±0.2 mm of their actual positions. Differences between daily EPID and Dynalog leaf-end positions were established and to be monitored on the long term. Monte Carlo simulation was used to recalculate the dose distribution of a given plan using EPID-extracted leaf positions and Dynalog files recorded leaf positions resulting in close dose volume histograms.