**Current research projects**

Below you can find a list with example projects and topics for MSc theses that we have available. Regardless of your past experience with medical imaging, we are quite certain you will find something that resonates. If you would have a specific topic in mind, besides the ones listed below, please get in touch with Prof. Jessica Bastiaansen. **Back to Research →**

MSc thesis project:

Building a neural network for accurate quantitative MRI maps

**Topic**: Neural networks, AI, deep learning, MRI data acquisition, signal processing, image reconstruction**Who**: Students with a background in computer science, electrical or bio(medical) engineering who are interested in the development of novel imaging applications that are directly translatable to the clinic.

**Background**: Parameter quantification from a complex MR signal profile under specific excitation schemes such as phase-cycled bSSFP has tremendous potential. Our current signal extraction models rely on signal matching using dictionaries, which became popular with the emergence of magnetic resonance fingerprinting (MRF) approaches. While the dictionary fitting methods are well suited for joint solutions, deep learning-based methods enjoy high inference speed and accuracy. In order to exploit both strengths, we are developing a deep learning method for the complex bSSFP signal profile to enable parameter mapping with the exploitation of joint information.**Project**: The prospective student will work with training data from the literature and MRI studies, inspired by MRI physics, build the proposed learning architecture and training it, and assess the finesse of the final product by employing simulation studies and *in vitro* experiments.

MSc thesis project:

Using deep-learning for water-fat fraction quantification

**Topic**: Deep learning, neural network, signal processing, dual water fat image reconstruction.**Who**: Students with a background in bio(medical) engineering, electrical engineering, math, or computer science who are interested in the development of novel reconstruction frameworks for water fat quantification.

**Background**: MRI acquisitions that use bSSFP enjoy the highest signal per unit of time. However, these acquired data are sensitive to off-resonance, including those arising from magnetic field inhomogeneities as well as molecular compounds that have a different chemical shift. Being able to extract the difference off-resonant sources from bSSFP data will enable the generation of high quality parametric water fat maps, that are free of image artifacts. However it is challenging to deal with signal asymmetries typically observed in bSSFP signal profiles that these off-resonance introduce. These asymmetries are not observed in conventional approaches for water fat fraction mapping such as GRE. We would like to combine the benefits of both approaches, by using neural networks trained on GRE data, to deal with off-resonance sensitivity observed in bSSFP.**Project**: The prospective student will construct a neural network that is trained on gold standard fat-fraction data to extract quantitative water fat maps from bSSFP data. The project will involve working with simulated data as well as real data that was acquired in the lab.

MSc thesis project:

Mathematical modeling in MRI for multi-parameter estimation

**Topic**: Mathematical modelling, group theory, MRI multi-parameter estimation, simulations, data processing and analysis, complex systems.**Who**: Students with a background in mathematics, physics, electrical or bio(medical) engineering. Everyone who is interested in creative mathematical modelling, simulations and parameter estimation on real experimental data.

**Background**: Mathematical models are essential for parameter estimation in quantitative MRI. The dynamics of complex MRI systems are based on different combinations of physical equations. Each physical equation by itself is well understood, but in complex combinations there are different possibilities for parameter estimation. The challenge is to find representations and theory frames which include the whole information about the complex system, while exhibiting lowest possible complexity for the highest possible model-robustness, i.e. irreducible representations, “occam’s razor”-principle. Often a system can be simplified by transforming the mathematical problem into a suitable representation e.g. polar coordinates, complex space, Fourier space or exploiting characterizable invariants of the system, group theory or other algebras. In finding those representations creativity, mathematical knowledge, and physical understanding of MRI is necessary.**Project**: In this project, the prospective student will extend or design mathematical models for parameter (T1, T2, water, fat) quantification using MRI signals extracted from phase cycled balanced steady state free precession (bSSFP) data. Instead of using conventional complex space representations, the student will investigate how the robustness of parameter estimation changes using representations such as quaternions and group theoretical approaches, which are not well understood. The model-robustness will be evaluated in simulations (Bloch-Simulations) as well as on experimental data.

MSc thesis project:

A physics-inspired journey through k-space

**Topic**: Machine learning, physics inspired RF pulse design, MRI simulations, trajectory optimization, fat suppression**Who**: Students with a BSc degree in Physics, Biomedical or Electrical Engineering**Background**: The suppression of fat signal is crucial for the visualization of small anatomical structures such as coronary arteries. It is also beneficial for reducing streaking artifacts that arise from static chest fat tissue as a result of undersampling and motion compensation. In our lab we have worked extensively on the development of novel RF pulse designs for water excitation, but its optimization is not trivial as many MRI acquisition parameters can influence the outcome. Moreover, the design of the k-space trajectory influences the type of signal contrast that is observed in the final MRI data. Therefore, RF pulse design and k-space trajectory induced signal weighting need to be optimized simultaneously. This calls for the development of smarter approaches that take into account the above constraints.

**Project**: The aim of the project is to design 3D k-space trajectories, in combination with the development of water-excitation pulses to obtain MR images with complete fat signal suppression. Initially the project involves simulations and RF pulse optimization, followed by data acquisition on MRI scanners, and image analysis.

MSc thesis project:

Solving MRI inversion problems for robust quantitative parameter extraction

**Topic**: Inversion problem, regularization, bSSFP, simulations, data processing and analysis, complex systems.**Who**: Students with a background in mathematics, computer science, physics, electrical or bio(medical) engineering. Everyone who is interested in new and creative mathematical models and solutions, regularization, computer simulations and parameter estimation on real experimental data.

**Background**: In quantitative MRI the dynamic of a complex physical system is described by the combination of a set of physical equations leading to a certain kind of quantitative model or quantitative map. Those respective models/maps are used for the quantification of e.g. water-fat fractions or T1- and T2-time. By trying to map acquired/simulated data to the quantitative parameter of interest, inversion problems are frequently encountered. Depending on the problem, inversions can be challenging to solve. The challenge can either originate on the sparsity of acquired data or also on the mathematical model itself. The quantification of multi-compartment systems in MRI exhibits a lot of inversion operations, where some problems arise as ambiguities or so called “ill-posed problems”. Sometimes it is possible to find solutions of the inversion by constraining the system to boundary conditions, by reducing the complexity or dimensionality of the model or bring it in another mathematical formulation. Some of those problems are not well understood yet and their solution is key for a robust quantification of multi-compartment systems in MRI.**Project**: In this project, it is aimed to first identify a system with known dynamics, and then device a solution technique. In particular, the prospective student will investigate inversion problems that are typical for parameter quantification in multi-compartment systems, like water and fat, with special focus on the bSSFP-sequence. The student will investigate whether a problem can be solved analytically/numerically by an unambiguous mathematical representation or develop a regularized inverse problem-solving technique. Solutions will be tested in simulations and on real experimental data.

**Development of dictionary matching algorithms for multi-echo GRE MRI data****Comparison of k-space trajectories for optimal MRI data acquisition and image reconstruction****Quantitative flow imaging with phase-cycled bSSFP**