Maximizing Performance Of Low-field MRI: Broadening Access To Healthcare Research In LMICs

Magnetic Resonance Imaging (MRI) is an invaluable tool for medicine and for health research, but the majority of the world’s population do not have access to it. Emerging low-field MRI systems offer a potential route to improve this: current efforts have focused on creating portable and lower cost devices which are open to customization. This ‘open science’ approach includes both software and hardware, empowering scientists and engineers in lower and middle income countries (LMICs) to customize and co-create technology for their own use. Much current innovation has come from reducing hardware cost, coupled with software that improves image quality retrospectively. This project will instead focus on designing new types of control system, capable of maximising data quality prospectively. We will use data obtained from external sensors to build an accurate ‘digital twin’ of the low-field scanner – this is a computer model which can predict its performance in any scenario. We will then use this digital twin along with methods from machine learning to control the system such that high quality images can be obtained under constrained conditions. We will demonstrate this by using this new control method to run a low-field MRI scanner only from battery power.