Marius Winter

• Semantic segmentation of human muscles in MR images
• Extraction of relevant muscle features (biomarkers)
• Investigation of the association of biomarkers with muscular changes and diseases

If you are interested in working on a thesis (BA/FA/MA) focused on topics such as feature selection/extraction, uncertainty estimation and explainability in machine learning, or related areas, we can arrange a meeting to discuss possibilities.

** Update 26. Nov:

I am offering a thesis in the field of Data Science in Medicine:
The project is based on a dataset comprising approximately 28,000 participants (NAKO dataset). Available data includes metadata (e.g., age, BMI, hand grip strength, lung function tests, etc.), whole-body MRI scans (two-point Dixon; neck-to-knee), and muscle characteristics extracted from the MR images (specifically the gluteus, psoas, and thigh muscle groups).

The goal of the thesis is to formulate interesting medical research questions and answer them using this relatively large dataset. To address the research questions, both machine learning methods and traditional statistical approaches can be employed.

Example research questions:

(a) Individuals with higher body mass must move greater amounts of mass. Therefore, do they also have greater muscle mass? Additionally, does their physical performance depend on the degree of fat in their muscles? Does higher occupational physical activity (e.g. craftsman vs. desk job) influence muscle volume and fat fraction?

In general, the focus here is on investigating the interplay between muscle characteristics, subjective assessments of physical fitness (collected via questionnaires), and objectively measured traits (e.g., hand grip strength). Note: It will be important to maintain a clear and structured narrative throughout the thesis.

(b) Using spirometry data (lung function tests), we can classify the severity of COPD. A structured approach can be followed, including: What is the problem? How many NAKO participants meet the criteria for COPD based on their spirometry data? What participant characteristics are associated with COPD? Are there specific links to our muscle metrics? Can insights gained from the analysis inform recommendations for COPD prevention?