Cardiac catheterisation is a common procedure in patients with congenital heart disease (CHD). During these procedures, a catheter and a guidewire are navigated into the cardiovascular system under fluoroscopic guidance, which is associated with significant radiation exposure and lack of soft tissue visualisation. Magnetic resonance imaging (MRI) is a promising alternative to fluoroscopy as it avoids ionising radiation, has excellent soft tissue contrast, and provide superior hemodynamic data. However, such interventional MRI system requires specialised MRI-compatible devices (catheters and guidewires) which currently have degraded mechanical properties. Modern low-field MRI scanners have recently emerged and represent an exciting avenue for interventional MRI where off-the-shelve standard devices can now be used safely. However, such scanners are associated with lower imaging framerate, signal-to-noise ratio (SNR), and image resolution.
The aim of this PhD is to enable MRI-guided cardiac catheterisation at low field by 1) developing an advanced online image reconstruction pipeline providing images with a high framerate, high SNR, and high spatial resolution required for real-time device tracking and 2) evaluating its potential during MRI-guided cardiac catheterisation in patients.
Aim 1/Year 1: Development of a deep learning-based image artifact suppression using a physics informed neural network to enable highly under-sampling acquisition and high frame rate imaging.
Aim 2/Year 2: Development of a deep-learning based image denoising approach for the reconstruction of high SNR images. Models including 2D, 2D+time, 2D+time+real-time motion correction will be evaluated.
Aim 3/Year 3: Development of deep learning based super resolution reconstruction to enhance image spatial resolution. Integration of temporal information and real-time motion correction will be explored.
Aim 4/Year 4: Clinical evaluation of the proposed technology in patients undergoing MRI-guided cardiac catheterisation.
Learning experience:
The student will develop a range of inter-disciplinary skills including in AI/deep learning, MRI physics, Medical imaging acquisition/reconstruction, computer programming, and cardiology.
