Study of an appropriate reconstruction algorithm for an innovative electron beam imaging system for dosimetry in IORT (intra operative radiation therapy)

Intra operative radiation therapy (IORT) is a technique based on delivery of a high dose of ionising radiation to the cancer tissue, after tumour ablation, during surgery. The Novac7 is a new linear accelerator expressly conceived for IORT that supplies electron beams at several energy, with high dose rate. These peculiar characteristics give rise to some complications with classical dosimetric techniques. In the framework of a research contract between ENEA and the Physics Department of Bologna, Italy, an original digital system has been developed to study and visualise the Novac7 electron beam in real time. The system is conceived as a grid consisting of two bundles of scintillating optical fibres (SOP) over-crossing each other, optically coupled with two arrays of photodiodes as read-out system. The problem of image reconstruction can be expressed as follows: there are only two profiles, or data arrays, which correspond respectively to the light emitted along the fibres perpendicular to the X-axis for the X-profile, and along the fibres perpendicular to the Y-axis for the Y-profile. This problem is not dissimilar to the reconstruction problem in tomography where several projections should be composed to trace them back to the original image. Unfortunately, here we have only two profiles: we have two one-dimensional profiles and are seeking one bidimensional image that could produce them. We selected a known beam image acquired by another digital instrument, then we extracted from it the two profiles simulating the acquisition arrays. Subsequently, we tested several reconstruction algorithms on these profiles, comparing the reconstructed image with the original one. We started from the simple sum algorithm until to iterative algorithms, searching the best compromise between the computational complexity and an high precision. We found that the iterative method is the best solution: it respects the geometrical characteristics and the absolute intensity values of the original image. Moreover it can reconstruct the image in a time of less than one second, a very good result.