Radiation dose mapping using magnetic resonance imaging in a superheated emulsion chamber

This work describes the magnetic resonance (MR) imaging techniques and image processing algorithms developed for radiation dosimetry with the superheated emulsion chamber. The chamber contains an emulsion of chloropentafluoroethane droplets in a tissue-equivalent glycerin-based gel. The droplets are highly superheated and expand into vapor bubbles upon exposure to irradiation. Brachytherapy sources can be inserted into the superheated emulsion chamber to create distributions of bubbles. The distribution of bubbles is then representative of the dose distribution to which the emulsion is exposed. Cumulating data from multiple independent exposures is required to calculate statistically significant bubble densities.;MR imaging is well suited to determining the bubble distribution. Susceptibility gradients at the interfaces between bubbles and gel are exploited to enhance contrast so microscopic bubbles can be imaged using relatively large voxel sizes. A conventional three-dimensional gradient echo imaging method is developed and applied to multiple independent irradiations of the superheated emulsion chamber from an 125I source. An image post-processing technique is developed to semi-automatically segment the bubbles from the images and to assess dose distributions based on the measured bubble densities. Relative bubble densities compare favorably to relative radial dose distributions calculated as recommended by Task Group 43 (TG43) of the American Association of Physicists in Medicine as well as Monte Carlo radiation transport simulations.;A three-dimensional, segmented, double sampled, echo-planar imaging (EPI) technique is subsequently developed and applied to an 125I source. Combining two-dimensional EPI with a conventional phase encode in the third dimension provides for rapid acquisition of susceptibility weighted images. Segmentation reduces artifacts produced by magnetic field inhomogeneities, while double sampling removes Nyquist ghosting. Post-processing is performed to segment the bubbles and to generate two-dimensional relative bubble density curves. Monte Carlo generated corrections are applied to convert bubble density to dose to water. The results are compared to TG43 using a distance to agreement map. It is shown that the superheated emulsions can be extended from performing point radiation dosimetry to generating one-dimensional and two-dimensional radiation dose maps.