Objectives
The student will work on the following tasks:
1. Analysis and restructuring of the existing pipeline
Review the current MRI-based workflow, understand and identify its major components, dependencies, inputs, and outputs.
2. Software-oriented modularization
Refactor the pipeline into a clearer and more maintainable structure, with separate modules for data input, water-fat separation, liver segmentation, feature generation, and prediction.
3. Development of a user-friendly interface
Design and implement a simple interface that allows non-technical users to run the workflow with minimal manual intervention. Depending on feasibility, this may take the form of a graphical interface, a guided command-line workflow, or a lightweight web-based front end.
4. Standardization of outputs and documentation
Define standardized outputs, including intermediate quality control information and final prediction results, and provide concise user documentation.
5. Prototype deployment and RAP feasibility assessment
Investigate how the software prototype can be deployed in a secure research environment such as RAP, including technical requirements, limitations, and possible implementation strategies.
6. Evaluation of the prototype
Perform an initial evaluation of the software with respect to reproducibility, usability, runtime behavior, and robustness in realistic usage scenarios.
Research component
In addition to software development, the project should include a scientific evaluation of the developed prototype. Possible research questions include:
7. How reproducible are the pipeline outputs across repeated runs and different execution settings?
8. Which parts of the workflow represent the main bottlenecks for usability and accessibility by non-technical users?
9. What are the practical challenges in translating a research-grade MRI analysis pipeline into a deployable software prototype for use in secure data environments?
This component is intended to ensure that the project is not only an implementation exercise, but also contributes methodological insight into research software development for medical AI workflows.
Expected Outcomes
The expected outcomes of the project include:
10. A modular prototype software for liver MRI-based analysis and prediction
11. A user-oriented workflow that can be applied by researchers without strong programming experience
12. Documentation of the software structure, inputs, and outputs
13. An initial assessment of deployment feasibility in RAP
14. A written evaluation of reproducibility, usability, and technical limitations
Preferred Skills
This project is suitable for a Master’s student with an interest in medical AI, biomedical data science, or research software engineering. Useful prior experience includes:
15. Python programming
16. Basic machine learning and deep learning knowledge
17. Interest in medical imaging or biomedical data analysis
18. Familiarity with software design, workflow automation, or interface development
Prior experience with MRI processing is helpful but not strictly required, provided the student is motivated to learn.