Introduction

At the Institute of Biomedical Engineering, a course focusing on implant manufacturing and testing was developed within the research field of implant technologies. In this course students focus on implant design, perform a parameter study based on a mathematical model, manually braid implants and finally test them. Over four years, continuous op-timizations allowed to improve implant quality and reproducibility. We retrospectively compare manufacturing alter-natives with regard to their applicability in educational perspective.

Methods

A microstructured mandrel was designed for manual braiding, whereas support rings ensure accurate shaping of im-plant loops. Braiding times of different students were measured and implant geometry was characterized. Different wire connection methods were mechanically evaluated and rated regarding simplicity and reproducibility. The influ-ence of heat treatment parameters on implant mechanics was examined. Finally, the reproducibility of the manufac-turing process and the calculations of the mathematical model were validated via mechanical tests.

Results

Braiding times to manufacture a functional implant varied under the unexperienced students and decreased within two days braiding to approximately two hours. Three connection methods were associated with considerably different me-chanical stability and behavior within a catheter. As expected, variation of heat treatment parameters led to differ-ences in implant mechanics. Accordingly, a suitable parameter set was selected. The reproducibility of the manufac-turing process was rated satisfactory with low variance in radial force. Delivery in an hydrogel aneurysm model using a 4 Fr catheter was feasible.

Conclusion

In the novel master course, students without previous expierence learned how to design and manufacture a braided implant in a time frame of six weeks. Next development steps are aimed at implant miniaturization and fluid dynamics evaluation.

In the bachelor program Medical Engineering, we have introduced the elective course Safety of Medical Electrical Equipment Lab Exercises. One of the lab exercises deals with the topic characteristics of different earthing systems and electricity supply networks in medical used areas. A laboratory setup has been developed and introduced for this lab exercise. With this setup it is possible to evaluate the characteristics of different earthing systems with extra-low voltage in order to avoid hazardous situations.

Due to the recent popularity and availability of Large Language Models (LLMs), creators of educational and training materials can more efficiently extract keywords for use in personalised learning recommendations than ever before. However, due to the LLMs' nontransparency in keyword extraction, an automation of this otherwise labour-intense task inherits the risk of biased content characterisation and recommendations. In this research, we present an original framework to enhance keyword selection through statistical analysis and generally available content metadata. For this, we collected 38 potential keywords (with their definitions) and 5 topics on dementia care from previous studies. To assess the algorithms support in enhanced keyword extraction, we then tasked 5 domain-experts and 3 general LLMs equipped with and without Retrieval Augmented Generation (RAG; for tailoring the models' generations to the content's domain) to extract keywords for each topic. Using Krippendorf's $\alpha$ metric, we then proceeded to select reliable keyword sets for inclusion and exclusion in content metadata, sensitive to the previous agreement amongst each rater group and topic. Last, we compared the LLM-based sets with those selected for humans to assess the support of each model for reliable keyword extraction. Overall, the results suggest that general LLMs struggle with the task (66\% of extractions were faulty for the first two models), and topic-wise internal agreement is low ( $\alpha$=0.59 (0.42) for model 3 (using RAG) on average; $\alpha$=0.68 for human raters). Due to this, the framework last resulted in a median set of 6 and 27 (9 and 24) out of 38 available keywords to be reliably included and excluded for each topic, with 67\% of keywords within the benchmark keyword sets selected for human raters being selected based on LLM extractions. To conclude, this approach shows effective in adapting to the content and extracted keywords presented.

It is recognized that teaching and embracing entrepreneurial skillsets lead to more valuable and impactful in- novations in biomedical engineering. Only very few institutions globally have started to integrate innovation generation methods and entrepreneurial mindset development in their curricula. A novel approach was developed and tested for the IEEE CAS society through a 6 week online training pro- gram leading the participants, accompanied by weekly personal guidance, to create a real innovation project. This was embedded in a competition to allow the best teams to receive funding to be able to attend the annual society conference and participate in a competition. The learners all reported a changed mindset towards innovation generation and the presented results were stimulating and exciting. The paper provides details on the free training program and shows some of the results. This could - and maybe should - become a standard training and program event point at future biomedical engineering conferences.