Development of Mathematical Phantom for MRI Simulators ［Presidential Award Proceedings］

Realistic numerical phantoms are required for accurate and appropriate MRI simulations. However, structure models used in conventional numerical phantoms are insufficient due to the limited number of components used for simulation. This study proposed an automated technique for the decomposition of brain structures for development of a numerical phantom with an arbitrary number of components. The acquired MRI scans of a volunteer's brain were decomposed into 207 components using a constrained linear least-squares method. Those components were approximated to 10 components using principal component analysis. The simulation and experimental studies demonstrated the usefulness of our method. The proposed approach will contribute to the development of a mathematical phantom, a numerical phantom defined with mathematical modeling.