Generating anthropomorphic phantoms semi-automatically from magnetic resonance images

The ultimate motivation of this work is to increase the accuracy of SPECT reconstructions by accounting for the types of patient motions that happen during acquisitions. Assessing the effectiveness of these motion correction algorithms on real patient motions is difficult, since there are currently no numerical anthropomorphic phantoms that accurately simulate non-periodic deformations such as bending or twisting. We address this problem by providing a way to generate anthropomorphic numerical phantoms of volunteers in different poses, which can be used to simulate imaging procedures in which the patient moves, as well as to assess the accuracy of motion correction algorithms.