Evolutionary MRI Magnet Design

Exhibitor: Bo Yuan

Supervisor: Marcus Gallagher

Research Group: Complex and Intelligent Systems

Industry Sector: Health / Medical Technology

Magnetic Resonance Imaging (MRI) is an imaging technique widely used in medical settings to produce high quality images of the inside of the human body. A typical system is like a cylinder with a deep bore in the centre where the patient is to be placed.

A key component of the MRI system is the magnet, which is used to generate an intense and homogeneous field in the region to be imaged(dsv). A superconducting magnet usually consists of a set of coils concentric about the Z axis and sysmmetric with regard to the X-Y plane. Each coil has a number of turns (superconducting wire) and has a rectangular cross-section.

A major challenge is to design short magnets (e.g., 1m) so that the perception of claustrophobia (a fear of enclosed places) experienced by patients undergoing MRI examinations can be reduced. This is a significant engineering challenge as the field homogeneity is strongly dependent on the overall length of the coil structure. Another challenge comes from the magnet design for special-purpose, open MRI systems where the imaging area is near the open end of the bore. The motivation is that, in some applications such as breast cancer detection, it would be much more convenient,compared to current whole-body systems,if there is a compact open MRI system that only requires part of the human body to be put into the system.

The parameters that need to be optimized in a magnet design include (for each coil):its position along Z-axis,its width along Z-axis,its inner radius and the number of turns. Furthermore, one more parameter is needed for each coil if we allow coils to be wound in both positive and negative directions.

This is a challenging optimization problem in that the overall search space is huge (i.e., consider playing with a number of rectangulars with variable sizes)and the goodness of a design is extremely sensitive to those parameters: a small change in one or more parameters can result in radical performance loss. In this project, we will apply several Evolutionary Algorithms (EAs),including but not limited to Genetic Algorithms(GAs), Differential Evolution (DE) and Estimation of Distribution Algorithms(EDAs). The ultimate objective is two fold. Firstly, we hope to find an efficient algorithm capable of designing magnets for short and/or open MRI systems. Secondly, this design problem can be also used as a test bed to evaluate various EAs.