Tom Strube

Mr. Tom Strube

orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 December 2024 in Current Directions in Biomedical Engineering
Verified
Journal Article
Open access
PDF file
TSTom Strube
Tom Strube
TNTom Nowak
Tom Nowak
MPMariia Pokotylo
Mariia Pokotylo
BKBernd Kuhlenkötter
Bernd Kuhlenkötter
Due to the recent popularity and availability of Large Language Models (LLMs), creators of educational materials can more efficiently extract keywords for use in personalised learning recommendations than ever before. However, due to the LLMs’ probabalistic nature, the automation of the otherwise labour-intense keyword extraction inherits the risk of biased and non-explainable results. In this research, we present an original framework to enhance keyword selection based on content title and description through a novel, reliability-sensitive, keyword selection algorithm. For this, we collected 38 potential keywords (together with their definitions) for five topics on dementia care from previous studies, together with two contents per topic. To assess the new method’s support in extracting keywords, we then prompted 5 human experts and 3 LLMs (using Retrieval Augmented Generation (RAG) for the keyword definitions) to select keywords to include and exclude for each content. Using Krippendorf’s a metric, we then were able to adapt to the present agreement, and to reliably select keyword sets for inclusion and exclusion for each content individually. Last, we compared these LLM-based keyword sets with those selected by humans to assess the impact of the adaptive keyword selection algorithm. Overall, the results suggest that LLMs generally struggle with the task (66% of extraction attempts either contained hallucinated or did not return any keywords), and topic-wise internal agreement is low ( a=0.59 (0.42) for model 3 (using RAG) on average; a=0.68 for human raters). Due to this, the reliable keyword selection resulted in a median set of 6|27 keywords for inclusion|exclusion per topic, with many of those keywords being within the benchmark keyword sets selected by human raters. To conclude, this approach shows effective in adapting to different levels of agreement in extracting keywords.


3 February 2023 in Energies
Verified
Journal Article
Open access
PDF file
KWKristin Wode
Kristin Wode
TSTom Strube
Tom Strube
ESEva Schischke
Eva Schischke
MHMarkus Hadam
Markus Hadam
SPSarah Pabst
Sarah Pabst
AMAnnedore Mittreiter
Annedore Mittreiter
Since existing energy system models often represent storage behavior in a simplified way, in this work, a tool chain for deriving consistent storage model parameters for optimization models is developed. The aim of our research work is to identify what are non-negligible influences on the the technical characteristics and dynamic behavior of the storage, to quantify the effect of these influences, and represent these effects in the model. This paper describes the developed tool chain and presents its application using an example. The tool chain consists of the steps “parameter screening”, “dynamic simulation”, “regression analysis” and “refining optimization model”. It is investigated which parameters have an influence on the storage system (here pumped hydroelectric energy storage (PHES)), how the storage behavior is modeled, which influencing factors have a measurable effect on the system, and how these findings can be integrated into optimization models. The main finding is that in the case of PHES, the dependency of the charging and discharging efficiency on the power is significant, but no further influencing factor has to be considered for accurate modeling (0.946 ≤ R2 ≤ 0.988) of the efficiency. It is concluded that the presented toolchain is suitable for other storage technologies as well, including the analysis of aging behavior.

21 May 2026
Verified
Book Chapter
Open access
TSTom Strube
Tom Strube
LWLeoni Weltermann
Leoni Weltermann
JWJonas Weber
Jonas Weber
JDJérôme Defosse
Jérôme Defosse
Artificial Intelligence (AI) and Clinical Practice Guidelines (CPGs) both aim to support clinical decision-making but may provide conflicting suggestions. This manuscript presents a Guideline-Aligned Machine Learning (GAML) model to predict ondansetron administration at the end of anaesthesia, based on Gan et al.’s Fourth Consensus Guidelines for the Management of Postoperative Nausea and Vomiting (PONV). n= 16,240 anaesthesia protocols were analysed for risk factors and administered PONV prophylaxes. Logistic regression, multinomial naïve Bayes, and CatBoost classifiers were trained on 80% of protocols with 12-fold cross-validation; optimal thresholds were set by the mean F1-maximising cut-off across folds. Models were evaluated on the remaining 20%, achieving high accuracy (90 ± 1%) and moderate precision and recall (60 ± 5%, 75 ± 4%) across all models. A SHAP decision plot was further computed on the test set to visualise predictor contributions and illustrate a potential interactive preoperative planning interface. Overall, GAML is a promising basis for explainable decision support in clinical care.


21 May 2026
Verified
Book Chapter
Open access
TSTom Strube
Tom Strube
LWLeoni Weltermann
Leoni Weltermann
JWJonas Weber
Jonas Weber
JDJérôme Defosse
Jérôme Defosse
Adoption of AI-based clinical decision support systems (AI-CDSS) remains limited, largely due to insufficient trust in AI models, especially in high-risk healthcare scenarios. To address this, this research presents an initial conceptual model linking explainability and reliability to clinician trust and intention to use AI-CDSS, thereby guiding the design of trustworthy AI-CDSS.