On scatter in megavoltage x-ray transmission imaging (in English)

This thesis considers the role of scatter in electronic portal imaging devices (EPIDs). The thesis involved a direct measurement of scatter in an EPID. An already established Monte Carlo code, developed at the Royal Marsden Hospital, Sutton/UK, was extended and agreed with the experimental data for phantom-to-detector distances larger than 10 cm. An analytical scatter model was developed on the basis of the Monte Carlo analysis. In this model scatter was estimated by convolving the input beam profile with an analytical scatter kernel. The model was tested over a wide range of clinical parameters. For a 6 MV x-ray beam impinging upon a water-equivalent phantom, the absolute error in the scatter-to-primary ratio was better than 1% for a 100 [formula omitted] field. The accuracy dropped to 4% for a 400 [formula omitted] field. The model was implemented for a camera-based EPID with a converter comprising a 1-mm-thick copper plate bonded to a 134 [formula omitted] phosphor foil. Algorithms were developed for correcting for scatter contributions in megavoltage CT (MVCT) and for reconstruction of input beam profiles. For the latter an overall accuracy of better than 4% was obtained for fields of 100 [formula omitted] in size. In MVCT the yield in contrast resolution after a scatter correction was small. However, a scatter correction reduced the discrepancy of 30% between true and reconstructed electron densities to 10%. In both applications the limit in accuracy is a direct consequence of the performance of the converter, which strongly overestimates multiply scattered photons of low energy. Within the framework of a convolution model multiply scattered radiation exiting an inhomogeneous phantom cannot be calculated with an accuracy exceeding the limits given. © 2000, American Association of Physicists in Medicine. All rights reserved.