PhD position on deep learning for quantitative MRI

Quantitative MRI is a PICNIK with Physics-Informed data Consistent Neural Implicit K-space representations. 

In this project, you will substantially improve quantitative magnetic resonance imaging (MRI) image quality using deep learning approaches. Quantitative MRI allows healthcare providers to quantitatively assess and characterize the state of a tumour and its microenvironment. This information can be used to personalize cancer treatments. For example, a well-perfused tumour (quantified with MRI) will be more likely to react to chemotherapy than a tumour that is not perfused, as the chemo will need to reach the tumour. As personalizing a treatment based on such biomarkers can substantially improve the efficacy of the treatment, we have multiple research lines that utilize quantitative MRI at Amsterdam UMC.

However, current quantitative MRI approaches have notoriously poor image quality, low resolution and poor precision. Consequently, quantitative MRI is not used routinely in cancer care. If we can improve the image quality of quantitative MRI, it can be used in clinical routine to select the optimal treatment for each patient, greatly improving treatment outcomes worldwide.

Therefore, in this PhD project, you will fundamentally change how quantitative MRI is obtained and develop innovative AI approaches for generating quantitative MRI images from raw MRI data. This will open the way to personalized treatments. These AI-driven frameworks will be tailored to suit the specific needs of quantitative MRI. In particular, we will use physics-informed neural-implicit k-space representations. By studying k-space (Fourier domain of the image in which the acquisition is performed) samples from over the entire time series, a neural-implicit representation can infer what the full k-space should look like at any given time. This way, we will achieve an image quality of quantitative MRI as if conventional MRI were being acquired. The project will balance between advanced deep learning methods and MRI reconstruction.

Your main task will be to implement, optimize and test new approaches to AI-driven quantitative MRI.

• You will work in our multi-disciplinary team, including AI-experts, MRI-physics experts, and clinical experts.
• You will conduct new academic research combining machine learning and medical physics.
• Your research will be published in reputable peer-reviewed journals. Additionally, you will regularly present your work at (international) conferences and assist in relevant teaching activities.

In this role, you will apply your technical expertise to improve healthcare. You are able to learn and understand the technical details of MRI acquisition and reconstruction and advanced deep learning. You must:

• have an MSc degree in Artificial intelligence, (Medical) Physics, or a similar study (please state your average grade and most relevant courses);
• have programming experience in Python;
• have English communication, presentation, and writing skills;
• have an interest in health care and, in particular, MRI;
• be highly motivated and creative.

Additionally, we value experience with machine learning (i.e. in PyTorch or Tensorflow) and experience with MRI and MR-physics.