1st Supervisor: Dr Sebastien Roujol, King’s College London
2nd Supervisor: Prof Amedeo Chiribiri, King’s College London
Clinical Supervisor: Prof Amedeo Chiribiri, King’s College London
Aim of the PhD Project:
- To develop novel magnetic resonance imaging (MRI) techniques for myocardial tissue characterisation for newly introduced low magnetic field MRI scanners for optimal accuracy, precision, and reproducibility
- To automate the workflow of these techniques (scan prescription, image acquisition/reconstruction, and post-processing/analysis) using AI
- To evaluate the proposed methods in phantom/healthy volunteers/patients
Lay Summary:
Cardiac magnetic resonance imaging (MRI) is commonly used for the clinical management of patients with cardiovascular diseases. cardiac MRI examinations are commonly performed using high magnetic field MRI scanners (1.5T/3T) which are expensive, and this has been an important limiting factor to their worldwide spread. Affordable low magnetic field (0.55T) MRI scanners have been recently developed as an alternative to break such barriers. The aim of this project will be to develop the next generation of time efficient cardiac MRI acquisition techniques for low field MRI scanners. This will involve the development of novel acquisition schemes taking advantage of specific properties of magnetic tissue properties at low field (such as shorter T1, longer T2, etc), novel acceleration techniques and AI-based reconstruction methods and to automate image analysis using AI to reduce operator dependency, analysis time, and cost.
The ideal candidate should have a strong interest in biomedical engineering, medical physics, computer science and related fields. Novel MRI acquisition/reconstruction techniques as well as new AI approaches will be developed in a truly interdisciplinary project to build a unique broad set of skills, currently in need in both the academic and industrial job market.
Figure 1. Example of myocardial T1 maps acquired using high field MRI scanners using our recently developed accelerated FAST1 approach (left) and conventional approach (right). A five-time acceleration rate was obtained without compromising accuracy and reproducibility of T1 maps.