Optical calorimetry (OC) is a novel form of radiation dosimetry that used two interfering beams of light to measure the radiation induced phase change in a volume of water, and relates the phase change to absorbed dose to water. It presents a promising dosimetric method for the measurement of FLASH therapy beams, where its measurement of dose to water, lack of components perturbing the radiation beam, and dose rate independence avoids many of the technical challenges facing traditional dosimetric methods in these ultra-high dose rate beams. The current OC dosimeter outputs a 2D image of dose, integrated across the volume of water. This study aims to investigate the requirements to accurately reconstruct a 3D dose distribution through performing a tomographic reconstruction upon the 2D OC dosimeter output.

Due to a desire to keep the mechanical complexity of a 3D OC dosimeter design minimised, this work looks at performing a single-projection reconstruction upon radially symmetric radiation beams using the inverse Abel transform, and performing a two-projection reconstruction on simple square and circular radiation beams using a filtered back projection or fast Fourier transform algorithm. A multi-projection reconstruction where the restriction on the number of projections used is removed was also performed, to indicate the level of mechanical complexity such an approach would entail if required.

Reconstruction results showed the single-projection inverse Abel transform method could reconstruct a manually integrated proton beam to an accuracy of better than 1%, but this error increased to 8% when using the projection data generated using the FRED virtual OC dosimeter model. This error could be reduced to below 4% when the level of noise in the FRED dosimeter model was reduced to 1/8th of the full amount, indicating that an accurate 3D reconstruction using a single projection is possible, but a substantial decrease in the noise within the detector system is required. The two-projection method performed accurately in some situations, did not reliably produce accurate reconstructions of simple photon beams with clinical features modelled. The multi- projection reconstruction investigation determined that at least 25 projections would be required to achieve a reconstruction with an error below 2% for all beams investigated, representing a significant increase in mechanical complexity of the OC dosimeter design.

The findings of this work recommend that in order to accurately reconstruct a 3D dose distribution using the OC dosimeter, the single-projection inverse Abel transform method could be used provided the noise in the dosimeter system is reduced, or the multi-projection FBP reconstruction method could be used provided that the additional mechanical complexity of such a design is accounted for.