Design of a Permanent Magnet Array for Head Imaging in a Low-Field Portable MRI System

In this thesis the problem of optimization of a permanent magnetic structure for low-field imaging applications is addressed, with particular attention to the homogeneity of the 𝐵𝑧 field within a spherical region of interest defined as FOV (Field of View). The whole work is based on the use of a numerical code in MATLAB environment, based on a Coulomb model for magnetic field calculation. In the initial phase, the code was subjected to an accurate validation by comparison with established software for magnetostatic analysis commonly used in academic and industrial environments. The good agreement between the results confirmed the reliability of the numerical approach adopted. Subsequently, several magnetic configurations from the literature have been analyzed to assess their performance in terms of intensity and homogeneity of the magnetic field generated in the FOV. Each configuration was simulated and compared systematically, with the aim of identifying the most promising geometry to be optimized. Among the options studied, one promising initial configuration was selected, which was then subjected to an extensive process of global optimization using a genetic algorithm (ga) to explore the space of possible geometries in depth. During the course of the work, the configuration has been progressively modified and enhanced, through the introduction of new structural elements and geometric variants to further improve its homogeneity and efficiency. With the proposed magnetic arrays, it was possible to obtain fields characterized by homogeneity of less than 25,000 ppm within a spherical FOV of 20 cm in diameter, with magnetic field intensity varying between 0.16 T and 0.19 T, depending on the specific changes made to the configuration. Although the results obtained represent a good goal in terms of homogeneity and intensity of the field,the current structure is not yet suitable for imaging applications since it lacks of gradient fields, essential for spatial encoding in MRI. Nevertheless, the configurations developed in this thesis provides provide a solid foundation for future developments towards more advanced 3D imaging solutions and integration with portable diagnostic systems.