List-mode proton CT reconstruction using their most likely paths via the finite Hilbert transform of the derivative of the backprojection
Résumé
Modern prototypes of proton computed tomography (CT) scanners can measure the energy, the position and the direction of each proton, before and after the scanned object, in a list mode. Each detected proton contributes to an estimate of a line integral along an estimated curved proton path. In this work, we propose a backproject first algorithm based on the two-step Hilbert transform to reconstruct proton CT images. The algorithm takes into account the estimated curved paths. A pixel-specific backprojection is computed from the average measurements of protons which traverse the pixel with the same direction according to the proton path estimates. Our simulations studies show that the algorithm has similar spatial resolution to a previous filtered backprojection (FBP) algorithm for proton CT using most likely paths while being computationally more efficient and able to handle truncated data.