Next generation magnetic resonance imaging through vision (2026-2029)
Abstract
Images acquired via magnetic resonance imaging (MRI) currently do not make any allowances for the way that humans read and understand regions of interest like segmenting important objects from the background. This project aims to unify artificial intelligence (AI) models for segmentation with MRI measurements and reconstruction directly in the frequency domain, helping us explain how such models work and create a new method of MRI image acquisition more akin to human vision that only acquires the areas an operator needs. The outcomes of the project have the potential to explain AI models, speed-up MRI scans and lower costs, thereby improving access to MRI services in the future and further advancing urgently needed AI research in Australia.
