ProModell – Projects


Spinal Cord Stimulation (SCS) is a therapeutic approach utilized for a range of conditions, including pain and movement disorders. However, the precise mechanisms underlying the effects of SCS remain mysterious, making clinical decisions often reliant on a trial-and-error methodology. To elucidate these mechanisms and enhance the precision of treatment, we are in the process of developing digital patient models of SCS. These models aim to deconstruct the underlying mechanisms, pave the way for innovative therapeutic strategies, and provide robust tools for clinical decision support. 

The primary challenge is that conventional MRI sequences often cannot identify the segmentation of spinal rootlets, leading to most segmentations being done manually. Although we have experimented with methods like Kalman filtering, the challenge persists. 

Potential topic directions

  • Mathematical Modeling for Physics
  • Computational Neuroscience
  • 3D CAD
  • Image-Based Reconstruction
  • Machine Learning & Deep Learning

Contact: Dr. Andreas Rowald

Research Projects (5-10 ECTS) or Thesis

Neural Root Segmentation and 3D Model Generation 

In this project, you will conduct in-depth research on a comprehensive T2 MRI dataset of the entire spinal cord, aiming to construct an efficient pipeline for automatic or semi-automatic neural root segmentation. By mapping MRI data to a standardized spinal cord space, you will gradually optimize the accurate segmentation of neural roots through a combination of manual and semi-automatic segmentation strategies. Utilizing the segmentation results, you will further explore innovative algorithms for generating 3D structures of neural roots. 

Our goal is to develop meticulously crafted spinal cord computational models that accurately incorporate neural root structures. This will provide more precise reference points for the advancement and application of neural stimulation technology in clinical settings. 

Requirements: Prior knowledge in image processing and MRI imaging. Bonus: Anatomy, 3D spatial path generation. 

Contact: Zhaoshun Hu

Application of Deep Learning in Neural Root Segmentation 

This project will focus on the field of deep learning, utilizing the SOTA deep learning networks to train multiple segmentation models on standard and raw spaces, as well as radial and axial images. The project consists of two key stages: firstly, participants will pre-train a segmentation network on T2 MRI spinal cord images, validating its segmentation performance on a comprehensive T2 MRI dataset. Subsequently, through transfer learning or network architecture redesign, we will apply this network to the domain of neural root segmentation. 

This project holds the potential to become a thesis topic for master’s research, primarily aiming to construct a 3D segmentation network for the automatic generation of neural roots. Simultaneously, we will explore generative networks based on segmentation data, aiming to pioneer novel methods for neural root generation. 

Requirements: Involvement in relevant deep learning courses, knowledge in MRI image processing. Bonus: Prior experience in deep learning 3D convolutions, familiarity with generative networks.

Contact: Zhaoshun Hu

Contact Information 

If you are interested in the above projects, please send an email to the contact. Please start the email subject with “[Application-Research Project + Project Name]”. Please attach your CV and academic transcript to the email (we recommend uploading the attachments to platforms like FAU box and sharing the link). We look forward to having you on board! 

Friedrich-Alexander-Universität Erlangen-Nürnberg