Studies in health technology and informatics最新文献

This chapter presents Reinforcement Learning (RL) based solutions for healthcare, highlighting its transformative potential across various domains. These studies commence by examining the utility of RL in precision medicine to showcase its ability to tailor treatment plans to individual patient profiles for optimal outcomes. The concept of a dynamic treatment regimen is discussed, demonstrating how RL algorithms can adjust therapies in response to a patient's evolving health status. It is also considered the applications of RL in home medication to demonstrate how intelligent systems can assist patients in managing their medications. Personalized rehabilitation is another critical area where RL algorithms facilitate customized rehabilitation protocols that enhance recovery rates and patient engagement. Adaptive healthcare interfaces powered by RL are explored, high-lighting their role in improving clinician-patient interactions and decision-making processes. The deployment of RL in diagnostic systems is examined, emphasizing improved diagnostic accuracy and early disease detection. The chapter discusses the use of RL in control systems as well, where it ensures the efficient operation of medical devices and systems, enhancing patient safety and treatment efficacy. Health management systems are also covered, indicating how RL can optimize resource allocation, workflow management, and patient monitoring to enhance overall healthcare delivery. The chapter concludes with a discussion of the limitations and potential future contributions of current RL applications in healthcare.

Available, high-quality CardioPulmonary Resuscitation (CPR, as a subset of Basic Life Support) training can reduce the impact of sudden cardiac arrests, a global health challenge. However, such training is labor intensive as instructors can only handle groups of 6 trainees, who have to judge their trainees performance via visual inspection. Consequently, both availability and quality of CPR training are under pressure. Current CPR manikins are either passive or are expensive and closed source, limiting their use significantly. We introduce a smart manikin with Technology-Enhanced Feedback (TEF). This smart TEF manikin can function autonomously, providing accurate, consistent, and real-time feedback, while being open source. It measures trainee's performance on CPR's three main metrics: ventilation volume, compression depth, and chest recoil. In due time, its Human-in-the-Loop (HITL) control framework will allow adaptive, personalized training, assessment of retention of training, and advanced life support scenarios. As such, it holds the promise for better CPR training and, hence, can save lives of those that suffer from an out-of-hospital cardiac arrest.

Major Depressive Disorder (MDD) stands as a leading cause of disability, demanding innovative self-management solutions. Despite advances in treatment and management, many still struggle with daily symptoms, highlighting the urgency for solutions that can offer scalable, personalized treatment to promote long-term well-being and reduce reliance on professional intervention. This study evaluates the effectiveness of BodyMindGlow, a digital platform designed for mental health self-management which records and monitors health status and provides recommendations to improve physical and mental well-being. App requirements were defined following a study involving 205 applications for mental health management, detailing current practices in Behavioral Management, Persuasive Systems Design (PSD), Behavioral Economics (BE) principles and Security and Privacy strategies. The study involved two groups: one with a clinical diagnosis of depression and/or anxiety and another without a diagnosis but interested in improving their general well-being. For two weeks, participants recorded mental and physical state daily and used services for emotional management and promotion of a healthy lifestyle. Follow-up was consulted daily through evolution graphs. Evaluation included initial assessment tests (PHQ-9 and GAD-7), daily records, a usability questionnaire divided into the System Usability Scale (SUS) and a personalized questionnaire. The results provided insights to guide future research and advancements in Digital Mental Health (DMH), demonstrating the effectiveness of an engaging solution that fits the fundamental and limiting needs of current mental health self-management. Creating this platform with advanced features, such as Artificial Intelligence, Gamification and SmartCoach, proved crucial to improve adherence and effectiveness of interventions.

Three-dimensional (3D) medical images are used for several purposes, including medical diagnosis and surgical planning. In both domains, eXtended Reality (XR) technologies are used to provide immersive working environments, where 3D pathological data can be analyzed. XR technologies include Virtual Reality (VR), Augmented Reality (AR), and Mixed Reality (MR). XR allows 3D imaging data to be viewed in a more realistic way compared to using 2D displays, reducing perceptual problems and improving spatial understanding. We report on a series of experimental studies in which XR technologies have been used to improve medical diagnosis or surgical planning. The results demonstrate that XR technologies can improve medical data understanding, for example, by providing efficient interaction techniques and clear visualizations.

Developments in robotics, among which exoskeletons, have lately known a constant increase. Wanting to understand how these devices transform realities of both able and impaired bodies, I conducted multi-sited ethnographic fieldwork in research centers where exoskeletons are designed and sometimes used. At one of these sites, I was allowed to take part in a test for an exoskeleton developed for persons with neurological impairments, being thus given the possibility to literally become my research phenomenon and thus enter the algorithm of a device. Data about my own able body contributed to forge forms of knowledge for experts in engineering sciences, but also to build an algorithm for a technological object designed for impaired people. I argue that my participation to my own object of study that resulted in my experiencing of a form of digital incorporation into this very object led me to theorize the category of "epistemological vulnerability" from a methodological point of view.

This book chapter focuses on the regulation of data altruism by organizations within the scope of "European Health Data Space" (EHDS) and "Data Governance Act" ("DGA") and turns to the organization of sharing rules when the personal data that finds the regulation of data subjects in the General Data Protection Regulation ("GDPR") becomes non-personal data. "Data Altruism Organization" (DAOs) are the organization of facilitating and strengthening the data sharing ecosystem and consent mechanism and systematize health improvement in the field of health data. In other words, explores the pivotal role of data subjects in the data economy, emphasizing the need for a framework that actively includes them to facilitate easier data utilization. As our understanding evolves, it becomes clear that strengthening the structure of "Data Altruism Organization" and data intermediary services is essential, particularly in light of "General Data Protection Regulation". GDPR"s regulation of relationships between data controllers and data subjects. It is crucial to address the economic and informational asymmetries that exist between these parties, recognizing that while data subjects elevate their data's status as a protected right, existing systems often limit responses to harm through restitution. Focusing on the secondary use of health data, this chapter, discusses advancements and objectives concerning the "Data Altruism Organization" outlined in the "Data Governance Act" ("DGA"). It also examines the interactions between data subjects and data controllers within the "European Health Data Space" (EHDS), noting that the "Data Governance Act" ("DGA") significantly enhances data protection through these altruism organizations. This governance framework aims to ensure the responsible use of data collected by public bodies for the public good while addressing challenges related to commercial confidentiality, intellectual property rights, and personal data. The "DGA" guarantees secure secondary use of health data through the framework of "Data Altruism Organization", with approved data intermediation service providers acting as reliable entities for appropriate data usage. The book chapter concludes by examining how this framework will be established and the operational dynamics of trustworthy data-sharing organizations in the health sector. Additionally, it clarifies the concept of data altruism as defined by the "DGA", highlighting its importance in promoting public interest objectives across various sectors, including healthcare and scientific research. It should also be said that the issue is examined only through the EU and the European Economic Area (EEA) and that the mere regulation does not concern the scope of Europe. Due to cross-border data sharing and the legality of sharing data with safe countries, European Regulations also concern third countries.

Artificial Intelligence (AI) is evolving at a rapid pace with impressive performance in many areas including health care. This study aimed to investigate the opportunities, challenges and barriers in implementing AI in patient monitoring using smart medical devices. This chapter provides a comprehensive analysis of the current state, potential applications, benefits, and challenges associated with the integration of AI into patient care settings. It could serve as a valuable resource for researchers, practitioners and policymakers interested in the transformative impact of AI on patient care. We reviewed studies that report on the opportunities and challenges in improving personalized patient care plans (PPCP) using AI. It is suggested that a holistic approach is required involving strategy, communications, integrations and collaboration between technology developers, healthcare professionals, regulatory bodies and end users including physicians and patients. Developing frameworks that prioritize ethical considerations, patient privacy, and model transparency is crucial for the responsible deployment of AI in healthcare. Balancing these opportunities and challenges requires collaboration between wider stakeholders to create a robust framework that maximizes the benefits of AI in healthcare while addressing the key challenges and barriers such as the explainability of the models, validation, regulation, and privacy integration with the existing clinical workflows.

As disasters become more frequent and severe, integrating smart health technologies into disaster preparedness has become essential for enhancing healthcare system resilience. This paper explores the role of technologies such as telemedicine, the Internet of Things (IoT), wearable devices, Electronic Health Records (EHRs), mobile health applications, and Artificial Intelligence (AI) in improving disaster management. These innovations enable real-time monitoring, remote healthcare delivery, and predictive analytics, which are crucial for mitigating the health impacts of disasters. The paper also addresses key challenges, including issues of interoperability, data security, and resistance to adoption. Case studies, such as AI's role in managing COVID-19 and IoT's application in natural disasters, demonstrate the effectiveness of these technologies. The research concludes by highlighting future directions, focusing on advancements in AI and IoT, as well as the importance of partnerships to overcome existing barriers and strengthen disaster preparedness.

Contemporary healthcare utilizes data stored as e-health data, personal health records (PHR), electronic health records (EHR), as well as data from wearables and sensors. These personal health (PH) data must be secure, private, and usable for analysis to improve patient treatments and overall healthcare quality. For practitioners, these data are essential; for governments, they support decision-making; and for society, they enable new medical discoveries. Given the sensitivity of PH data, ensuring data confidentiality is crucial. Data must be standardized, accurate, and timely to be reliable for medical use. Key challenges include security, privacy, consent management, and legal compliance. Different legal and technical measures can be used to address these challenges. In this study, we consider data security and privacy from the whole aspect of the healthcare data life cycle, as well as the most important laws that regulate healthcare data security and privacy, starting from the treats to the solutions published in the literature. Privacy-preserving techniques are continually advancing, with significant developments in trusted execution environments and cryptographic methods. Current best practices involve strict adherence to consent and privacy policies, ensuring that individuals' data is handled with the utmost care. NoPeek learning techniques allow for data analysis without sharing the actual data, thereby enhancing privacy. This approach should be combined with differential privacy, a technique that adds statistical noise to data to prevent the identification of individual data points. Artificial Intelligence (AI) promises advancements in healthcare but requires adherence to regulations. Disease prediction models using AI analyze vast datasets to predict health-related threats but must balance benefits with data protection and regulatory compliance. Collaborative approaches can integrate predictive analytics into personalized healthcare while maintaining trust and ethical standards.

Background: Menopause, endometriosis, miscarriage, and female infertility are health issues affecting women worldwide (nearly half the global population). The biomedical literature is human-readable and evergrowing with around 3.5K papers published daily. Current Artificial Intelligence (AI) cannot reliably deal with facts, and automatic processing of scientific publications to obtain reliable insights remains a challenge, although, representing diseases in an actionable (machine-interpretable semantics) has a long-standing tradition in biomedical research.

Objective: Conduct some experiments to explore to what extent ontologies and knowledge graphs (symbolic AI) can support comprehensive human-centric explainable AI for artificial neural networks (neural AI).

Methods: Instead of using Large Language Models (LLMs) from neural AI as all-in-one generative AI solution, this paper investigates a neuro-symbolic AI approach, combining neural AI (to process and extract patterns for health issues from free-text) with symbolic AI (explicit representations of background knowledge). Our neuro-symbolic AI approach leverages on domain knowledge (simple classification based on predefined categories), and scientific evidence from the biomedical literature to provide human-readable explanations (explainable AI) formally represented as nanopublications (machine-processable knowledge graphs). We investigated if incorporating prior domain knowledge (best scientific evidence) into vector arithmetic formulas may support customisation (e.g. bringing "unseeing" terms for a predefined category).

Results: We performed three experiments (EXP1, EXP2 and EXP3), evaluating 315 candidate n-grams obtained by applying unsupervised vector arithmetic formulas (cosine for similarity and 3CosAdd for four-term analogies) to word2vec embeddings created from 301,201 PubMed citations (titles and abstracts). We also conducted a fourth experiment (EXP4) with 9 LLMs to automatically extract and classify terms from evidence-based text excerpts, evaluating 381 terms from LLMs' output. In EXP4, we looked into the output categories provided by 2 open-source small-size biomedical LLMs (with 66.4 and 184 millions of trainable parameters) and 7 free-of-charge general LLMs (DeepSeek-V3, Groq, Grok3-beta, QWEN2.5-MAX, Gemini, Claude, ChatGPT4).

Conclusion: Biomedical knowledge may guide explainability (what predictions from neural models are worthy to explain and what predictions can be ignored) and enable a higher level of customisation when using word2vec embeddings and LLMs for extracting patterns for health issues from free-text.