The generation of synthetic data holds significant promise for augmenting limited datasets while avoiding privacy issues, facilitating research, and enhancing machine learning models’ robustness. Generative Adversarial Networks (GANs) stand out as promising tools, employing two neural networks—generator and discriminator—to produce synthetic data that mirrors real data distributions. This study evaluates GAN variants (CTGAN, CopulaGAN), a variational autoencoder, and copulas on diverse real datasets of different complexity encompassing numerical and categorical attributes. The results highlight CTGAN’s sensitivity to training parameters and TVAE’s robustness across datasets. Scalability challenges persist, with GANs demanding substantial computational resources. TVAE stands out for its high utility across all datasets, even for high-dimensional data, though it incurs higher privacy risks, which is indicative of the curse of dimensionality. While no single model universally excels, understanding the trade-offs and leveraging model strengths can significantly enhance synthetic data generation (SDG). Future research should focus on adaptive learning mechanisms, scalability enhancements, and standardized evaluation metrics to advance SDG methods effectively. Addressing these challenges will foster broader adoption and application of synthetic data.

In view of the various methodological developments regarding the protection of sensitive data, especially with respect to privacy-preserving computation and federated learning, a conceptual categorization and comparison between various methods stemming from different fields is often desired. More concretely, it is important to provide guidance for the practice, which lacks an overview over suitable approaches for certain scenarios, whether it is differential privacy for interactive queries,k-anonymity methods and synthetic data generation for data publishing, or secure federated analysis for multiparty computation without sharing the data itself. Here, we provide an overview based on central criteria describing a context for privacy-preserving data handling, which allows informed decisions in view of the many alternatives. Besides guiding the practice, this categorization of concepts and methods is destined as a step towards a comprehensive ontology for anonymization. We emphasize throughout the paper that there is no panacea and that context matters.

Background: Nurse scheduling is a complex challenge in health care, impacting both patient care quality and nurse well-being. Traditional scheduling methods often fail to consider individual preferences, leading to dissatisfaction, burnout, and high turnover. Inadequate scheduling practices, including restricted autonomy and lack of transparency, can further reduce nurse morale and negatively affect patient outcomes. Research suggests that participative scheduling approaches incorporating nurse preferences can improve job satisfaction. Artificial intelligence (AI) and mathematical optimization methods, such as mixed-integer programming (MIP), constraint programming (CP), genetic programming (GP), and reinforcement learning (RL), offer potential solutions to optimize scheduling and address these challenges. Objective: This study aims to develop a framework for integrating nurses’ preferences into AI-supported scheduling methods by gathering qualitative insights from nurses and supervisors and mapping these to mathematical and AI-based scheduling techniques. Methods: Focus group interviews were conducted with 21 participants (nurses, supervisors, and temporary staff) from Swiss health care institutions to understand experiences and preferences related to staff scheduling. Qualitative data were analyzed using open and axial coding to extract key themes. These themes were then mapped to AI methodologies, including MIP, CP, GP, and RL, based on their suitability to address identified scheduling challenges. Results: The study revealed key priorities in nurse scheduling. Fairness and participation were highlighted by 85% (18/21) of interview participants, emphasizing the need for transparent and inclusive scheduling. Flexibility and autonomy were preferred by 76% (16/21), favoring shift swaps and self-scheduling. AI expectations were mixed: 62% (13/21) saw potential for improved efficiency and fairness, while 38% (8/21) expressed concerns over reliability and loss of human oversight. Mapping to AI methods showed MIP as effective for fair shift allocation, CP for complex rule-based conditions, GP for handling unforeseen absences, and RL for dynamic schedule adaptation in hospital environments. A preliminary AI implementation of MIP in a training hospital unit (35 staff members) showed how to design a system from a mathematical perspective. Conclusions: AI-supported scheduling systems can significantly enhance fairness, transparency, and efficiency in nurse scheduling. However, concerns regarding AI reliability, adaptability to individual needs, and human oversight must be addressed. A hybrid approach integrating AI recommendations with human decision-making may be optimal. Future research should explore the broader implementation of AI-driven scheduling models and assess their impact on nurse satisfaction and patient outcomes over time.

The rising demand to use genetic data for research goes hand in hand with an increased awareness of privacy issues related to its use. Using human genetic data in a legally compliant way requires an examination of the legal basis as well as an assessment of potential disclosure risks. Focusing on the relevant legal framework in the European Union, we discuss open questions and uncertainties around the handling of genetic data in research, which can result in the introduction of unnecessary hurdles for data sharing. First, we discuss defining features and relative disclosure risks of some DNA-related biomarkers, distinguishing between the risk for disclosure of (1) the identity of an individual, (2) information about an individual's health and behavior, including previously unknown phenotypes, and (3) information about an individual's blood relatives. Second, we discuss the European legal framework applicable to the use of DNA-related biomarkers in research, the implications of including both inherited and acquired traits in the legal definition, as well as the issue of “genetic exceptionalism”—the notion that genetic information has inherent characteristics that require different considerations than other health and medical information. Finally, by mapping the legal to specific technical definitions, we draw some initial conclusions concerning how sensitive different types of “genetic data” may actually be. We argue that whole genome sequences may justifiably be considered “exceptional” and require special protection, whereas other genetic data that do not fulfill the same criteria should be treated in a similar manner to other clinical data. This kind of differentiation should be reflected by the law and/or other governance frameworks as well as agreed Codes of Conduct when using the term “genetic data.”

Natural Language Processing (NLP) has shown promise in fields like radiology for converting unstructured into structured data, but acquiring suitable datasets poses several challenges, including privacy concerns. Specifically, we aim to utilize Large Language Models (LLMs) to extract medical information from dialogues between ambulance staff and patients to populate emergency protocol forms. However, we currently lack dialogues with known content that can serve as a gold standard for an evaluation. We designed a pipeline using the quantized LLM “Zephyr-7b-beta” for initial dialogue generation, followed by refinement and translation using OpenAI’s GPT-4 Turbo. The MIMIC-IV database provided relevant medical data. The evaluation involved accuracy assessment via Retrieval-Augmented Generation (RAG) and sentiment analysis using multilingual models. Initial results showed a high accuracy of 94% with “Zephyr-7b-beta,” slightly decreasing to 87% after refinement with GPT-4 Turbo. Sentiment analysis indicated a qualitative shift towards more positive sentiment post-refinement. These findings highlight the potential and challenges of using LLMs for generating synthetic medical dialogues, informing future NLP system development in healthcare.

Background: Intraoperative neurophysiological monitoring (IONM) guides the surgeon in ensuring motor pathway integrity during high-risk neurosurgical and orthopedic procedures. Although motor-evoked potentials (MEPs) are valuable for predicting motor outcomes, the key features of predictive signals are not well understood, and standardized warning criteria are lacking. Developing a muscle identification prediction model could increase patient safety while allowing the exploration of relevant features for the task. Objective: The aim of this study is to expand the development of machine learning (ML) methods for muscle classification and evaluate them in a bicentric setup. Further, we aim to identify key features of MEP signals that contribute to accurate muscle classification using explainable artificial intelligence (XAI) techniques. Methods: This study used ML and deep learning models, specifically random forest (RF) classifiers and convolutional neural networks (CNNs), to classify MEP signals from routine supratentorial neurosurgical procedures from two medical centers according to muscle identity of four muscles (extensor digitorum, abductor pollicis brevis, tibialis anterior, and abductor hallucis). The algorithms were trained and validated on a total of 36,992 MEPs from 151 surgeries in one center, and they were tested on 24,298 MEPs from 58 surgeries from the other center. Depending on the algorithm, time-series, feature-engineered, and time-frequency representations of the MEP data were used. XAI techniques, specifically Shapley Additive Explanation (SHAP) values and gradient class activation maps (Grad-CAM), were implemented to identify important signal features. Results: High classification accuracy was achieved with the RF classifier, reaching 87.9% accuracy on the validation set and 80% accuracy on the test set. The 1D- and 2D-CNNs demonstrated comparably strong performance. Our XAI findings indicate that frequency components and peak latencies are crucial for accurate MEP classification, providing insights that could inform intraoperative warning criteria. Conclusions: This study demonstrates the effectiveness of ML techniques and the importance of XAI in enhancing trust in and reliability of artificial intelligence–driven IONM applications. Further, it may help to identify new intrinsic features of MEP signals so far overlooked in conventional warning criteria. By reducing the risk of muscle mislabeling and by providing the basis for possible new warning criteria, this study may help to increase patient safety during surgical procedures.