Frontiers in digital health最新文献

{"title":"Artificial intelligence and health empowerment in rural communities and landslide- or avalanche-isolated contexts: real case at a fictitious location.","authors":"Rune Johan Krumsvik, Vegard Slettvoll","doi":"10.3389/fdgth.2025.1655154","DOIUrl":"10.3389/fdgth.2025.1655154","url":null,"abstract":"<p><p>Through a series of case studies, we have pretested the capabilities and reliability of the Large Language Models (LLM), Generative Pre-trained Transformer 4 (GPT-4) and OpenAI o3 reasoning model (o3) in educational and healthcare contexts. Based on this knowledge, we took a step further by testing these technologies in an authentic patient case set in a fictitious location. The context for this brief case report relates to the fact that, in the first quarter of 2025, fewer patients lacked an assigned GP compared to previous years-a positive trend. However, this offers little relief to those cut off from GP care due to their rural location or because of landslides and extreme weather. This case highlights the need for knowledge-based preparedness and alternative health empowerment pathways in rural Norway. This brief case report describes a single 16-year-old boy (<i>N</i> = 1) with no significant past medical history or chronic conditions. Although he lived in an urban area, we reframed the encounter as a simulated rural, avalanche-isolated scenario to test the feasibility of AI-supported care under extreme access constraints. Specifically, the case models how a patient in an avalanche-prone mountain valley-where seasonal road closures routinely sever access to healthcare facilities-could receive rapid, guideline-concordant treatment for severe tonsillitis during a period of general-practitioner (GP) unavailability. Repeated attempts to secure a same-day appointment were thwarted by workforce shortages and impassable roads, resulting in the earliest available appointment being five days away. The family leveraged point-of-care technologies (fingerstick C-reactive protein analysis, wearable sensors, blood pressure device, digital fever device, mobile ECG) and an o3 language model[1] to evaluate disease severity. A peak CRP of 130 mg/L, combined with otherwise stable vital signs, prompted a remote consultation with a trusted physician in their social network, who confirmed the diagnosis of bacterial tonsillitis and initiated treatment with phenoxymethylpenicillin (Apocillin). Within 72 h, CRP fell to 23 mg/L and symptoms were resolved. The patient case and the events described in this pilot study are authentic, but the location is fictitious. The waiting time to see a general practitioner was five days in both the actual urban setting and the simulated rural scenario; however, unlike in urban contexts-where patients can often access immediate care through emergency clinics or private GPs-such options are typically unavailable in sparsely populated rural areas. This case illustrates how AI and health technology can serve as a \"virtual waiting room\" for individuals in rural or landslide- and avalanche-isolated areas, especially when GP access is limited and the condition is low-risk, such as mild sore throat symptoms. The case illustrates how inexpensive diagnostics and AI-supported reasoning can strengthen health empowerment and temporarily b","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"7 ","pages":"1655154"},"PeriodicalIF":3.2,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12417106/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145042429","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Generative AI in consumer health: leveraging large language models for health literacy and clinical safety with a digital health framework.","authors":"Annemarie K Tilton, Brian E Caplan, Brian J Cole","doi":"10.3389/fdgth.2025.1616488","DOIUrl":"10.3389/fdgth.2025.1616488","url":null,"abstract":"<p><p>Generative AI, powered by large language models, is transforming consumer health by enhancing health literacy and delivering personalized health education. However, ensuring clinical safety and effectiveness requires a robust digital health framework to address risks like misinformation and inequitable communication. This mini review examines current use cases for generative AI in consumer health education, highlights persistent challenges, and proposes a clinician-informed framework to evaluate safety, usability, and effectiveness. The RECAP model-Relevance, Evidence-based, Clarity, Adaptability, and Precision-offers a pragmatic lens to guide responsible implementation of AI in patient-facing tools. By connecting insights from past digital health innovations to the opportunities and pitfalls of large language models, this paper provides both context and direction for future development.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"7 ","pages":"1616488"},"PeriodicalIF":3.2,"publicationDate":"2025-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12417475/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145042546","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Performance of vision language models for optic disc swelling identification on fundus photographs.","authors":"Kelvin Zhenghao Li, Tuyet Thao Nguyen, Heather E Moss","doi":"10.3389/fdgth.2025.1660887","DOIUrl":"10.3389/fdgth.2025.1660887","url":null,"abstract":"<p><strong>Introduction: </strong>Vision language models (VLMs) combine image analysis capabilities with large language models (LLMs). Because of their multimodal capabilities, VLMs offer a clinical advantage over image classification models for the diagnosis of optic disc swelling by allowing a consideration of clinical context. In this study, we compare the performance of non-specialty-trained VLMs with different prompts in the classification of optic disc swelling on fundus photographs.</p><p><strong>Methods: </strong>A diagnostic test accuracy study was conducted utilizing an open-sourced dataset. Five different prompts (increasing in context) were used with each of five different VLMs (Llama 3.2-vision, LLaVA-Med, LLaVA, GPT-4o, and DeepSeek-4V), resulting in 25 prompt-model pairs. The performance of VLMs in classifying photographs with and without optic disc swelling was measured using Youden's index (YI), F1 score, and accuracy rate.</p><p><strong>Results: </strong>A total of 779 images of normal optic discs and 295 images of swollen discs were obtained from an open-source image database. Among the 25 prompt-model pairs, valid response rates ranged from 7.8% to 100% (median 93.6%). Diagnostic performance ranged from YI: 0.00 to 0.231 (median 0.042), F1 score: 0.00 to 0.716 (median 0.401), and accuracy rate: 27.5 to 70.5% (median 58.8%). The best-performing prompt-model pair was GPT-4o with role-playing with Chain-of-Thought and few-shot prompting. On average, Llama 3.2-vision performed the best (average YI across prompts 0.181). There was no consistent relationship between the amount of information given in the prompt and the model performance.</p><p><strong>Conclusions: </strong>Non-specialty-trained VLMs could classify photographs of swollen and normal optic discs better than chance, with performance varying by model. Increasing prompt complexity did not consistently improve performance. Specialty-specific VLMs may be necessary to improve ophthalmic image analysis performance.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"7 ","pages":"1660887"},"PeriodicalIF":3.2,"publicationDate":"2025-08-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12415036/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145031205","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Designing for dignity: ethics of AI surveillance in older adult care.","authors":"Jeena Joseph","doi":"10.3389/fdgth.2025.1643238","DOIUrl":"10.3389/fdgth.2025.1643238","url":null,"abstract":"","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"7 ","pages":"1643238"},"PeriodicalIF":3.2,"publicationDate":"2025-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12411420/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145016786","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Simulation-based assessment of digital twin systems for immunisation.","authors":"Leonardo de Oliveira El-Warrak, Claudio Miceli de Farias, Victor Hugo Dias Macedo De Azevedo Costa","doi":"10.3389/fdgth.2025.1603550","DOIUrl":"10.3389/fdgth.2025.1603550","url":null,"abstract":"<p><strong>Background: </strong>This paper presents the application of simulation to assess the functionality of a proposed Digital Twin (DT) architecture for immunisation services in primary healthcare centres. The solution is based on Industry 4.0 concepts and technologies, such as IoT, machine learning, and cloud computing, and adheres to the ISO 23247 standard.</p><p><strong>Methods: </strong>The system modelling is carried out using the Unified Modelling Language (UML) to define the workflows and processes involved, including vaccine storage temperature monitoring and population vaccination status tracking. The proposed architecture is structured into four domains: observable elements/entities, data collection and device control, digital twin platform, and user domain. To validate the system's performance and feasibility, simulations are conducted using SimPy, enabling the evaluation of its response under various operational scenarios.</p><p><strong>Results: </strong>The system facilitates the storage, monitoring, and visualisation of data related to the thermal conditions of ice-lined refrigerators (ILR) and thermal boxes. Additionally, it analyses patient vaccination coverage based on the official immunisation schedule. The key benefits include optimising vaccine storage conditions, reducing dose wastage, continuously monitoring immunisation coverage, and supporting strategic vaccination planning.</p><p><strong>Conclusion: </strong>The paper discusses the future impacts of this approach on immunisation management and its scalability for diverse public health contexts. By leveraging advanced technologies and simulation, this digital twin framework aims to improve the performance and overall impact of immunization services.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"7 ","pages":"1603550"},"PeriodicalIF":3.2,"publicationDate":"2025-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12411477/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145016789","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Heartbeat detection and personal authentication using a 60 GHz Doppler sensor.","authors":"Takuma Asano, Shintaro Izumi, Hiroshi Kawaguchi","doi":"10.3389/fdgth.2025.1570144","DOIUrl":"10.3389/fdgth.2025.1570144","url":null,"abstract":"<p><strong>Background: </strong>Microwave Doppler sensors, capable of detecting minute physiological movements, enable the measurement of biometric information, such as walking patterns, heart rate, and respiration. Unlike fingerprint and facial recognition systems, they offer authentication without physical contact or privacy concerns. This study focuses on non-contact seismocardiography using microwave Doppler sensors and aims to apply this technology for biometric authentication.</p><p><strong>Method: </strong>We proposed a method for authenticating and identifying heartbeat signals through supervised learning using a conditional variational autoencoder (CVAE). A 60 GHz microwave Doppler sensor was used to capture heartbeat signals, which were processed using a conformer network to detect peaks and segment individual beats. High signal-to-noise ratio waveforms were selected, and time-frequency analysis extracted relevant features. Spectrograms labeled with subject data were input into the CVAE, which encoded subject-specific features into a latent space for authentication.</p><p><strong>Results: </strong>The proposed heartbeat-based authentication method, validated on 13 subjects, achieved an average balanced accuracy of 97.3% for authentication and an average accuracy of 94.7% for identification. Compared with conventional methods, this approach demonstrated superior performance by effectively encoding subject-specific features while mitigating noise-related challenges.</p><p><strong>Conclusion: </strong>The proposed method enhanced the feasibility of non-contact heartbeat-based authentication by achieving high accuracy while addressing noise-related challenges. Its application could improve biometric security without compromising user privacy. Further advancements in handling posture variations and scalability are essential for real-world implementation.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"7 ","pages":"1570144"},"PeriodicalIF":3.2,"publicationDate":"2025-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12408575/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145016751","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Biases in AI: acknowledging and addressing the inevitable ethical issues.","authors":"Bjørn Hofmann","doi":"10.3389/fdgth.2025.1614105","DOIUrl":"10.3389/fdgth.2025.1614105","url":null,"abstract":"<p><p>Biases in artificial intelligence (AI) systems pose a range of ethical issues. The myriads of biases in AI systems are briefly reviewed and divided in three main categories: input bias, system bias, and application bias. These biases pose a series of basic ethical challenges: injustice, bad output/outcome, loss of autonomy, transformation of basic concepts and values, and erosion of accountability. A review of the many ways to identify, measure, and mitigate these biases reveals commendable efforts to avoid or reduce bias; however, it also highlights the persistence of unresolved biases. Residual and undetected biases present epistemic challenges with substantial ethical implications. The article further investigates whether the general principles, checklists, guidelines, frameworks, or regulations of AI ethics could address the identified ethical issues with bias. Unfortunately, the depth and diversity of these challenges often exceed the capabilities of existing approaches. Consequently, the article suggests that we must acknowledge and accept some residual ethical issues related to biases in AI systems. By utilizing insights from ethics and moral psychology, we can better navigate this landscape. To maximize the benefits and minimize the harms of biases in AI, it is imperative to identify and mitigate existing biases and remain transparent about the consequences of those we cannot eliminate. This necessitates close collaboration between scientists and ethicists.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"7 ","pages":"1614105"},"PeriodicalIF":3.2,"publicationDate":"2025-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12405166/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145002061","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Challenges of identification and anonymity in time-continuous data from medical environments.","authors":"Freimut Hammer, Thorsten Strufe","doi":"10.3389/fdgth.2025.1604001","DOIUrl":"10.3389/fdgth.2025.1604001","url":null,"abstract":"<p><p>In medical environments, time-continuous data, such as electrocardiographic records, necessitates a distinct approach to anonymization due to the paramount importance of preserving its spatio-temporal integrity for optimal utility. A wide array of data types, characterized by their high sensitivity to the patient's well-being and their substantial interest to researchers, are generated. A significant proportion of this data may be of interest to researchers beyond the original purposes for which it was collected. This necessity underscores the pressing need for effective anonymization methods, a challenge that existing approaches often fail to adequately address. Robust privacy mechanisms are essential to uphold patient rights and ensure informed consent, particularly within the framework of the European Health Data Space. This paper explores the challenges and opportunities inherent in developing a novel approach to anonymize such data and devise suitable metrics to assess the efficacy of anonymization. One promising approach is the adoption of differential privacy to account for temporal context and correlations, making it suitable for time-continuous data.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"7 ","pages":"1604001"},"PeriodicalIF":3.2,"publicationDate":"2025-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12405402/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145002058","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The Noor Project: fair transformer transfer learning for autism spectrum disorder recognition from speech.","authors":"Najla D Al Futaisi, Björn W Schuller, Fabien Ringeval, Maja Pantic","doi":"10.3389/fdgth.2025.1274675","DOIUrl":"10.3389/fdgth.2025.1274675","url":null,"abstract":"<p><p>Early detection is crucial for managing incurable disorders, particularly autism spectrum disorder (ASD). Unfortunately, a considerable number of individuals with ASD receive a late diagnosis or remain undiagnosed. Speech holds a critical role in ASD, as a significant number of affected individuals experience speech impairments or remain non-verbal. To address this, we use speech analysis for automatic ASD recognition in children by classifying their speech as either autistic or typically developing. However, due to the lack of large labelled datasets, we leverage two smaller datasets to explore deep transfer learning methods. We investigate two fine-tuning approaches: (1) Discriminative Fine-Tuning (D-FT), which is pre-trained on a related dataset before being tuned on a similar task, and (2) Wav2Vec 2.0 Fine-Tuning (W2V2-FT), which leverages self-supervised speech representations pre-trained on a larger, unrelated dataset. We perform two distinct classification tasks: (a) a binary task to determine typicality, classifying speech as either that of a typically developing (TD) child or an atypically developing (AD) child; and (b) a four-class diagnosis task, which further classifies atypical cases into ASD, dysphasia (DYS), or pervasive developmental disorder-not otherwise specified (NOS), alongside TD. This research aims to improve early recognition strategies, particularly for individuals with ASD. The findings suggest that transfer learning methods can be a valuable tool for autism recognition from speech. For the typicality classification task (TD vs. AD), the D-FT model achieved the highest test UAR (94.8%), outperforming W2V2-FT (91.5%). In the diagnosis task (TD, ASD, DYS, NOS), D-FT also demonstrated superior performance (60.9% UAR) compared to W2V2-FT (54.3%). These results highlight the potential of transfer learning for speech-based ASD recognition and underscore the challenges of multi-class classification with limited labeled data.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"7 ","pages":"1274675"},"PeriodicalIF":3.2,"publicationDate":"2025-08-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12399525/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144994669","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Preparing hospitals and health organizations for AI: practical guidelines for the required infrastructure.","authors":"Emil Byberg, Marco Crimi","doi":"10.3389/fdgth.2025.1605006","DOIUrl":"10.3389/fdgth.2025.1605006","url":null,"abstract":"","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"7 ","pages":"1605006"},"PeriodicalIF":3.2,"publicationDate":"2025-08-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12400513/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144994639","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}