Evaluation of Speech Quality Using Synthetic Data and Transfer Learning

[Type: Master Thesis, Focus: Medical, Machine Learning, Signal Processing]

​

This project explores the use of advanced AI methods for the automated assessment of speech quality in patients undergoing MR-guided Focused Ultrasound (MRgFUS) treatment. By combining synthetic data generation and transfer learning, the goal is to develop a robust and clinically applicable solution that can be integrated into existing signal-processing pipelines. The project is a collaborative effort between the chair of Digital Signal Processing and System

Theory (DSS), Visual Computing and Artificial Intelligence (VCAI), and the University Hospital Schleswig-Holstein (UKSH), bridging applied AI research and real-world clinical practice. The full thesis proposal is available here (link).

​

AI-based Implant Modeling

[Type: Bachelor or Master Thesis, Focus: Computer Vision, Medical, Machine Learning]

​

The Institute for Digital Implant Research (IDIR) is investigating how digital twins can be used to improve the development of patient-specific implants. This includes the sub-projects aneurysms, bone nails and heart valves. They all have the same approach in common: We aim to create a digital twin from real data (e.g., CT, MRI) of the patient. The optimum implant for the patient must now be determined/developed by simulation, which then only needs to be manufactured and implanted in the final step. In this context, there are always exciting questions that could be dealt with in the course of a Bachelor's or Master's thesis. If you are interested, please get in touch and we will see if we can find a suitable topic.

​

Bone Strength Analysis (FemoraLyze)

[Type: Bachelor or Master Thesis, Focus: Computer Vision, Medical, Machine Learning]

​

FemoraLyze is a modular framework for a fully-automated proximal femur analysis. Based on different segmentation masks, the framework is able to collect metrics such as cortical thickness in different bone areas, volumes, angles and distances as well as bone structure parameters. These are relevant in medical prognosis, diagnosis and therapy. In order to further increase the benefits for medical professionals, this framework is to be expanded with additional modules (e.g., finite element analysis) and modalities (e.g., MRI or clinical CT). Runtime optimization and a GUI would also be desirable.

​

Prediction of Bone Healing Processes using Sparse Longitudinal Data

[Type: Master Thesis, Focus: Computer Vision, Medical, Machine Learning]

​

The aim of this project is to investigate whether bone healing processes can be predicted on the basis of sparse longitudinal data, both in terms of overall extent and temporal resolution. To this end, a dummy CT data set must first be created, on which the various AI models will then be trained and tested.​​​​​​

​

Super-Resolution through Contrastive Alignment of Low- and High-Resolution Image Representations

[Type: Bachelor or Master Thesis, Focus: Super-Resolution, Deep Learning, Generative AI]

We are interested in exploring a CLIP-based approach to image super-resolution. The main idea is to learn a shared embedding space for paired low-resolution and high-resolution images by training separate encoders whose feature representations are aligned using a contrastive objective. Building on this shared latent space, the goal is to develop a model that reconstructs high-resolution images from the learned representation. Current areas of interest include the design of contrastive training objectives for paired image data, reconstruction from shared embedding spaces, and the comparison with established super-resolution approaches such as SRGAN, Real-ESRGAN, LDM, and ResShift. We are also interested in evaluating such methods on medical imaging data, where super-resolution may be particularly valuable for improving image quality in sensitive application domains.

​

Learning Artifact-Free Counterfactuals for Artifact Detection in Medical Imaging

[Type: Bachelor or Master Thesis, Focus: Diffusion Model, Deep Learning, Generative AI]

We are interested in exploring a diffusion-based approach for artifact detection in medical imaging, with a focus on CT, MRI, and ultrasound. Inspired by Synomaly Noise and Multi-Stage Diffusion (Link) , the core idea is to train a diffusion model exclusively on artifact-free medical images so that it learns the distribution of clean acquisitions. During inference, only images containing artifacts are provided to the model in order to generate counterfactual artifact-free reconstructions. By comparing the corrupted input with the reconstructed image, a residual image can be obtained in which the artifact should become visible. A natural extension of the Synomaly idea is to replace synthetic pathological anomalies during training with synthetic artifact corruptions that mimic modality-specific acquisition effects. This would allow the model to learn artifact removal in an unsupervised manner while preserving the underlying anatomy. Current areas of interest include the design of realistic artifact simulations, counterfactual reconstruction using diffusion models, and the localization and categorization of artifacts from residual maps. Possible evaluation settings include artifact detection in MRI, CT, and ultrasound, as well as comparisons across different artifact types and imaging modalities.​