Self-driving fetal MRI

About the Project

Magnetic Resonance Imaging (MRI) provides exceptional insight into anatomy and development without using radiation. Each MRI examination consists of several sequences, each designed to capture different tissue contrasts or orientations. However, when the subject moves during acquisition, image quality degrades and sequences must often be repeated, extending scan time. This problem is particularly severe in fetal MRI, where fetal motion and maternal breathing cause frequent image artefacts, and where the unpredictable position of the fetus requires careful, manual planning by experienced operators.

Previous work demonstrated that artificial intelligence can enable the scanner to automatically track and follow the fetus in real time during MRI. This PhD project will extend that capability toward a self-driving fetal MRI examination, capable of autonomously planning and adapting the scan as it runs. The project will focus on developing methods for dynamic slice prescription, adaptive sequence control, and real-time adjustment of imaging parameters in response to motion and anatomy. By integrating these approaches into the scanner workflow, the system aims to acquire complete, high-quality data efficiently and reliably—minimising operator input and improving both patient comfort and throughput.