MRI evaluation and image processing in gel dosimetry. A study of selected MRI properties and image processing in 3D gel dosimetry

Sammanfattning: Gel dosimetry is a new dosimetry method applied in radiation therapy. Gel dosimeters consist of a radiation sensitive gel, which can integrate absorbed doses from several radiation sources or beams. The dose to the gel can be evaluated by magnetic resonance imaging (MRI), a procedure that is the focus of the present thesis. A robust tool for the evaluation of the nonuniformity in MRI has been developed, the Deviation Image method. Unlike previously presented methods, the Deviation Image method includes all nonuniformity variations across a phantom surface and is insensitive to stochastic noise. Methods for the estimation of stochastic noise were analyzed in terms of sensitivity to nonuniformities. A method that averages the stochastic noise level over five regions over the phantom surface, and a method that assesses the stochastic noise level from the background, were found to be the methods of choice. Pronounced MR image nonuniformity variation with repetition and T1 relaxation time was observed in the spin-echo T1 measurement protocol. These variations were caused by nonuniform RF transmission in combination with the inherent differences in the allowance of T1 relaxation for different repetition times. Neither the T1 calculation itself, the uniformity optimized repetition times, nor the correction methods could sufficiently correct for these nonuniformities. The nonuniformities were found to vary considerably less with inversion time for the inversion-recovery pulse sequence, resulting in a T1 image with considerably lower nonuniformity. A multi-spin-echo pulse sequence using the 3D volume acquisition technique was developed that was capable of evaluating polyacrylamide gel dosimeters with an equal resolution of 1 mm in all three spatial dimensions. Expected advantages for the 3D technique in favor of the 2D techniques, could not be achieved in the actual measurements. Further development and studies of the 3D technique are therefore required, prior to its application to polyacrylamide gel dosimetry. An image processing computer software was developed, intended to integrate the image processing for the three-dimensional gel dosimetry treatment plan verification. The resulting PMRelax image processing software was found to be capable of performing the necessary image processing tasks. The image processing tasks were further associated with a graphical user interface in order to support routine gel dosimetry use. Two image-processing methods for spatial registering and relocation of absorbed dose distributions in two and three dimensions were studied. One method used inherent absorbed dose information as references and the other external references. Using the absorbed dose reference method, uncertainties associated with the spatial registering and relocation method were found difficult to predict, which warranted a need for external reference. Using fiducial markers, an accurate spatial registering and relocation in three dimensions was performed. Influence from spatial registering and relocation uncertainties could thereby be reduced as a source of error for the absorbed dose distributions obtained with gel dosimeters and treatment planning systems.