J. Mark Ansermino

Mahdi Yousefi, Klaske van Heusden, Ian M. Mitchell et al.

This work provides a formalized model-invariant safety system for closed-loop anesthesia that uses feedback from measured data for model falsification to reduce conservatism. The safety system maintains predicted propofol plasma concentrations, as well as the patient's blood pressure, within safety bounds despite uncertainty in patient responses to propofol. Model-invariant formal verification is used to formalize the safety system. This technique requires a multi-model description of model-uncertainty. Model-invariant verification considers all possible dynamics of an uncertain system, and the resulting safety system may be conservative for systems that do not exhibit the worst-case dynamical response. In this work, we employ model falsification to reduce conservatism of the model-invariant safety system. Members of a model set that characterizes model- uncertainty are falsified if discrepancy between predictions of those models and measured responses of the uncertain system is established, thereby reducing model uncertainty. We show that including falsification in a model-invariant safety system reduces conservatism of the safety system.

Aditya Nagori, Ayush Gautam, Matthew O. Wiens et al.

Clustering patient subgroups is essential for personalized care and efficient resource use. Traditional clustering methods struggle with high-dimensional, heterogeneous healthcare data and lack contextual understanding. This study evaluates Large Language Model (LLM) based clustering against classical methods using a pediatric sepsis dataset from a low-income country (LIC), containing 2,686 records with 28 numerical and 119 categorical variables. Patient records were serialized into text with and without a clustering objective. Embeddings were generated using quantized LLAMA 3.1 8B, DeepSeek-R1-Distill-Llama-8B with low-rank adaptation(LoRA), and Stella-En-400M-V5 models. K-means clustering was applied to these embeddings. Classical comparisons included K-Medoids clustering on UMAP and FAMD-reduced mixed data. Silhouette scores and statistical tests evaluated cluster quality and distinctiveness. Stella-En-400M-V5 achieved the highest Silhouette Score (0.86). LLAMA 3.1 8B with the clustering objective performed better with higher number of clusters, identifying subgroups with distinct nutritional, clinical, and socioeconomic profiles. LLM-based methods outperformed classical techniques by capturing richer context and prioritizing key features. These results highlight potential of LLMs for contextual phenotyping and informed decision-making in resource-limited settings.