Thesis and Project Guidelines

Please read the following guidelines carefully if you are planning to conduct a final thesis or research project (e.g., the Computational Imaging Project) at the Computational Imaging Lab!

  • If you are interested in doing your thesis with us, please apply using the application form below. To ensure a fair and organized process, we kindly ask that all supervision requests go through this form.
  • We encourage you to submit your own topic, related to our lab research. General topics and fields of research can typically be found on the personal pages of the individual group members.
  • Use the application form to write a text where you formulate a research question and provide an outline of your chosen approach. Please include 1-2 paper references as supporting literature for your approach.
  • For most topics, students should complete the Computational MRI course (lecture and exercise) beforehand, as it provides essential skills to work on your topic.

We encourage students to propose their own research topics based on their interests. A great starting point is exploring the research fields of our lab members to find inspiration. If you do not yet have a specific topic in mind, you may request one of our topics. However, we strongly prefer students to take the initiative in shaping their own research questions. We look forward to your ideas and enthusiasm!

Research Profile – Annika Hofmann

My work focuses on developing advanced MRI sequences and post-processing techniques to study the complex microstructure of brain tissue. The concept behind this work is called Q-Space Trajectory Imaging (QTI), which expands the conventional Diffusion Tensor Imaging (DTI) by using higher order tensors to capture anisotropic diffusion within a voxel. Unlike DTI, which models diffusion using a second-order tensor, QTI incorporates fourth-order tensors, to calculate additional anisotropy and coherence metrics, such as macroscopic anisothropy (𝐶𝑀), microscopic anisotropy (𝐶𝜇), coherence (𝐶𝐶), size variance measures (𝐶𝑀𝐷), and microscopic fractional anisotropy (𝜇FA). Sequence programming is done using IDEA (Siemens Software) as well as PyPulseq.
Images acquired with these sequences require post-processing. In this field we focus on motion correction as well as top-up correction and B0-Field correction. All operations are done in image domain after the reconstruction of the diffusion weighted images.
Beyond Sequence development and post-processing we also work on acquisition protocol optimization. A large focus of this work is on optimizing the diffusion gradient shapes and directions with toolboxes like NOW.

Research Profile – Erik Gösche

My research focuses on developing deep learning (DL) models to improve MRI image reconstruction, making the process faster and more accurate. MRI scans take time because they collect a lot of data, and speeding up this process can reduce patient discomfort and increase efficiency. My main research area is dynamic contrast-enhanced MRI (DCE-MRI), a technique that tracks how contrast agents move through tissues over time. DCE-MRI is widely used in cancer and cardiovascular imaging, but its effectiveness is often limited by slow acquisition times. Deep learning offers a powerful solution by learning patterns from large datasets, allowing us to reconstruct high-quality images even from limited data.

My work involves training and optimizing neural networks specifically for MRI reconstruction. I focus on unrolled end-to-end models like MoDL and VarNet, which simulate how traditional optimization algorithms work but in a learned, data-driven way. As ground-truth images are not always available for training I work with self-supervised techniques like SSDU, which do not require such data. Additionally, I explore novel architectures such as transformers, dual-domain networks, and coordinate-based networks to further enhance MRI reconstruction. In addition to image reconstruction, I am also interested in MRI segmentation, which involves identifying and outlining specific structures within the images. Segmentation plays a key role in medical image analysis, helping with tasks like tumor detection and organ localization.

Thesis Registration

The thesis has to be registered in the examinations office. This should be done soon after you have started working on your topic. For students of Medical Engineering, please review the [homepage of the study program] for the latest version of the registration form and fill it together with your supervisor. Students of other study programs coordinate directly with their supervisor regarding the registration procedure.

Lab Meetings

The CIL staff team meets every week for a lab meeting. After certain milestones in your project or thesis, your supervisor will encourage you to join the meeting to give a short presentation (≈15 min) to all lab members.

MRI Colloquium

The MRI colloquium is hosted together with other labs working on MRI and takes place on Thursdays at 11am during the lecture period. We encourage you to attend the colloquium regularly.

Presentation

The results of your thesis or project should be presented in a 30-minute final presentation in the MRI colloquium or the lab meeting. This presentation should be given 2-4 weeks before your deadline. This gives you the chance to include additional aspects from the discussion in your thesis or report. Please discuss the date of your presentation with your supervisor as early as possible to make sure that your favoured date is still available. The contribution of the presentation to your final grade depends on your study program.

Code

All of your code is required to be archived in a Git repository in our group on the [GitLab server of the RRZE]. Your supervisor will create a repository for you and invite you to it. We recommend to get familiar with Git as early as possible and to use it right from the beginning of your implementation work. Please find more information about the GitLab server [here]. Access to the GitLab server needs to be requested via the [IdM-Portal]. Note that Git is intended for text-based files (i.e., source code), and not for large files like PDFs or images. For this reason, projects on [gitlab.rrze.fau.de] have a quota of 2GB.

Language

The common language at the lab is English. Hence, all oral presentations should be given in English. Master’s theses should be written in English. Bachelor’s theses may be written in German, if your supervisor agrees. However, we highly recommend to write it in English as well.

Writing your Thesis / Research Project

The thesis or report needs to be written with LaTeX using the CIL thesis [template]. Your thesis should follow the common structure of scientific documents. Please also consider style, syntax, grammar, and punctuation as this highly influences the readability of your work and will also be taken into account for your grade. Please also refer to the guidelines of the Pattern Recognition Lab [PR Thesis Guidelines] when preparing your thesis.

Thesis Submission

Your thesis needs to be submitted as one printout version and as a PDF file to your supervisor before your deadline. Your implementations have to be handed over to your supervisor in the form of a Git repository on the [GitLab server of the RRZE]. This includes all code that is required to reproduce your results. All data that you have used or generated for your thesis needs to be accessible to your supervisor after your submission. This includes all your results, scripts for processing the data (if not already included in your Git repository), figures, etc. Therefore, please upload such material to our [thesis archive]. Details on the submission format of this data may vary and should be discussed individually with your supervisor. In any way, it is your responsibility to provide your supervisor with everything that would be required to reproduce your results.

  •