Hasil untuk "Computer applications to medicine. Medical informatics"

S. Sternberg

Li Wang, Xi Chen, Xiangwen Deng et al.

The use of large language models (LLMs) in clinical medicine is currently thriving. Effectively transferring LLMs’ pertinent theoretical knowledge from computer science to their application in clinical medicine is crucial. Prompt engineering has shown potential as an effective method in this regard. To explore the application of prompt engineering in LLMs and to examine the reliability of LLMs, different styles of prompts were designed and used to ask different LLMs about their agreement with the American Academy of Orthopedic Surgeons (AAOS) osteoarthritis (OA) evidence-based guidelines. Each question was asked 5 times. We compared the consistency of the findings with guidelines across different evidence levels for different prompts and assessed the reliability of different prompts by asking the same question 5 times. gpt-4-Web with ROT prompting had the highest overall consistency (62.9%) and a significant performance for strong recommendations, with a total consistency of 77.5%. The reliability of the different LLMs for different prompts was not stable (Fleiss kappa ranged from −0.002 to 0.984). This study revealed that different prompts had variable effects across various models, and the gpt-4-Web with ROT prompt was the most consistent. An appropriate prompt could improve the accuracy of responses to professional medical questions.

Nimisha Schneider, Paul Fabian, Michelle Cawley et al.

Ulugbek Shernazarov, Rostislav Svitsov, Bin Shi

Fine-tuning large language models for domain-specific tasks such as medical text summarization demands substantial computational resources. Parameter-efficient fine-tuning (PEFT) methods offer promising alternatives by updating only a small fraction of parameters. This paper compares three adaptation approaches-Low-Rank Adaptation (LoRA), Prompt Tuning, and Full Fine-Tuning-across the Flan-T5 model family on the PubMed medical summarization dataset. Through experiments with multiple random seeds, we demonstrate that LoRA consistently outperforms full fine-tuning, achieving 43.52 +/- 0.18 ROUGE-1 on Flan-T5-Large with only 0.6% trainable parameters compared to 40.67 +/- 0.21 for full fine-tuning. Sensitivity analyses examine the impact of LoRA rank and prompt token count. Our findings suggest the low-rank constraint provides beneficial regularization, challenging assumptions about the necessity of full parameter updates. Code is available at https://github.com/eracoding/llm-medical-summarization

Tayyaba Ali, Sidra Iqbal

Extensive parental use of electronic devices correlates with poorer parent-adolescent interactions, though research has not investigated any potential effects on adolescent behavior. This research investigated whether increased technoference is associated with higher levels of adolescents' internalizing and externalizing behaviors, along with diminished prosocial behaviors. 450 pakistani adolescents from public and private schools aged 11–17 completed the self-reported versions of The Technoference Scale and the Strengths and Difficulties Questionnaire. Results indicated that parental and adolescent technoference was positively correlated with internalizing and externalizing behavior problems, while negatively correlated with prosocial behavior. Strong association between parental and adolescent technoference was observed. Findings from this study highlight the significant influence of technoference on adolescent behavior, suggesting that managing technology within families is essential for promoting healthier behavioral patterns. The significant correlations between technoference and both internalizing and externalizing behaviors underscore the potential risks associated with excessive media use and disrupted family interactions.

Yixiong Chen, Xue Zhang, Sheng Zhang et al.

Abstract BackgroundHand, foot, and mouth disease (HFMD) is a global health concern requiring a risk assessment framework based on systematic factors analysis for prevention and control. ObjectiveThis study aims to construct a comprehensive HFMD risk assessment framework by integrating multisource data, including historical incidence information, environmental parameters, and web-based search behavior data, to improve predictive performance. MethodsWe integrated multisource data (HFMD cases, meteorology, air pollution, Baidu Index, and public health measures) from Bao’an District of Shenzhen city in Southern China (2014‐2023). Correlation analysis was used to assess the associations between HFMD incidence and systematic factors. The impacts of environmental factors were analyzed using the Distributed Lag Nonlinear Model. Seasonal Autoregressive Integrated Moving Average model and advanced machine learning methods were used to predict HFMD 1-4 weeks ahead. Risk levels for the 1- to 4-week-ahead forecasts were determined by comparing the predicted weekly incidence against predefined thresholds. ResultsFrom 2014 to 2023, Bao’an District reported a total of 118,826 cases of HFMD. Environmental and search behavior factors (excluding sulfur dioxide) were significantly associated with HFMD incidence in nonlinear patterns. For 1-week-ahead prediction, Seasonal Autoregressive Integrated Moving Average using case data alone performed best (RrRr ConclusionsThe epidemic dynamics of HFMD are influenced by multiple factors in a nonlinear manner. Integrating multisource data, particularly web-based search behavior, significantly enhances the accuracy of short- and midterm forecasts and risk assessment. This approach offers practical insights for developing digital surveillance and early warning systems in public health.

Jian Song, Binyu Yang, Mayur Desai et al.

Aim To summarise the facilitators and barriers influencing the utilisation of mobile health (mHealth) for rehabilitation among older adults with hip fractures, family caregivers, and healthcare providers. Methods A total of five databases (PubMed, Cochrane Library, Embase, Web of Science Core Collection, and Ovid) were searched from inception to July 2025. Additionally, grey literature and reference lists were also searched. Publications were eligible if they reported on facilitators or barriers influencing the utilisation of mHealth by older adults with hip fractures, family caregivers, or healthcare providers. Results Nine articles were included. Our findings indicated that the influencing factors included patient-related and mHealth technology-related factors. Both patients and family caregivers recognised mHealth's advantages in communication with healthcare providers, finding it useful and efficient, and effective in delivering rehabilitation instructions. Healthcare providers emphasised its value in delivering holistic care, providing health education, and facilitating patients and family caregivers’ engagement in disease management. However, patients and family caregivers reported both preferences for traditional healthcare models and functional limitations as key barriers. All groups jointly reported adoption barriers, including patients’ inadequate technical literacy, patients’ characteristics, lack of critical resources, as well as technology-related challenges concerning external constraints. Conclusion The study reveals the unique characteristics of older adults with hip fractures have constrained the deeper implementation of mHealth. Future research should target family caregivers and healthcare providers by systematically examining the critical factors influencing their use of mHealth in managing older adults with hip fractures. Such investigations would enhance the quality of clinical care, facilitate patient recovery, and improve prognostic outcomes.

Bailiang Jian, Jiazhen Pan, Rohit Jena et al.

Medical image registration drives quantitative analysis across organs, modalities, and patient populations. Recent deep learning methods often combine low-level "trend-driven" computational blocks from computer vision, such as large-kernel CNNs, Transformers, and state-space models, with high-level registration-specific designs like motion pyramids, correlation layers, and iterative refinement. Yet, their relative contributions remain unclear and entangled. This raises a central question: should future advances in registration focus on importing generic architectural trends or on refining domain-specific design principles? Through a modular framework spanning brain, lung, cardiac, and abdominal registration, we systematically disentangle the influence of these two paradigms. Our evaluation reveals that low-level "trend-driven" computational blocks offer only marginal or inconsistent gains, while high-level registration-specific designs consistently deliver more accurate, smoother, and more robust deformations. These domain priors significantly elevate the performance of a standard U-Net baseline, far more than variants incorporating "trend-driven" blocks, achieving an average relative improvement of $\sim3\%$. All models and experiments are released within a transparent, modular benchmark that enables plug-and-play comparison for new architectures and registration tasks (https://github.com/BailiangJ/rethink-reg). This dynamic and extensible platform establishes a common ground for reproducible and fair evaluation, inviting the community to isolate genuine methodological contributions from domain priors. Our findings advocate a shift in research emphasis: from following architectural trends to embracing domain-specific design principles as the true drivers of progress in learning-based medical image registration.

Johannes Kiechle, Stefan M. Fischer, Daniel M. Lang et al.

The sharp rise in medical tomography examinations has created a demand for automated systems that can reliably extract informative features for downstream tasks such as tumor characterization. Although 3D volumes contain richer information than individual slices, effective 3D classification remains difficult: volumetric data encode complex spatial dependencies, and the scarcity of large-scale 3D datasets has constrained progress toward 3D foundation models. As a result, many recent approaches rely on 2D vision foundation models trained on natural images, repurposing them as feature extractors for medical scans with surprisingly strong performance. Despite their practical success, current methods that apply 2D foundation models to 3D scans via slice-based decomposition remain fundamentally limited. Standard slicing along axial, sagittal, and coronal planes often fails to capture the true spatial extent of a structure when its orientation does not align with these canonical views. More critically, most approaches aggregate slice features independently, ignoring the underlying 3D geometry and losing spatial coherence across slices. To overcome these limitations, we propose TomoGraphView, a novel framework that integrates omnidirectional volume slicing with spherical graph-based feature aggregation. Instead of restricting the model to axial, sagittal, or coronal planes, our method samples both canonical and non-canonical cross-sections generated from uniformly distributed points on a sphere enclosing the volume. We publicly share our accessible code base at http://github.com/compai-lab/2025-MedIA-kiechle and provide a user-friendly library for omnidirectional volume slicing at https://pypi.org/project/OmniSlicer.

Md. Fazlul Karim Khondakar, Md. Hasib Sarowar, Mehdi Hasan Chowdhury et al.

Abstract Neuromarketing is an emerging research field that aims to understand consumers’ decision-making processes when choosing which product to buy. This information is highly sought after by businesses looking to improve their marketing strategies by understanding what leaves a positive or negative impression on consumers. It has the potential to revolutionize the marketing industry by enabling companies to offer engaging experiences, create more effective advertisements, avoid the wrong marketing strategies, and ultimately save millions of dollars for businesses. Therefore, good documentation is necessary to capture the current research situation in this vital sector. In this article, we present a systematic review of EEG-based Neuromarketing. We aim to shed light on the research trends, technical scopes, and potential opportunities in this field. We reviewed recent publications from valid databases and divided the popular research topics in Neuromarketing into five clusters to present the current research trend in this field. We also discuss the brain regions that are activated when making purchase decisions and their relevance to Neuromarketing applications. The article provides appropriate illustrations of marketing stimuli that can elicit authentic impressions from consumers' minds, the techniques used to process and analyze recorded brain data, and the current strategies employed to interpret the data. Finally, we offer recommendations to upcoming researchers to help them investigate the possibilities in this area more efficiently in the future.

Janis M Miller, Jean F Wyman, Lawrence An et al.

BackgroundAlthough surveys and apps are available for women to report urination and bladder symptoms, they do not include their decisions regarding toileting. Real-world factors can interfere with toileting decisions, which may then influence bladder health. This premise lacks data per want of a robust data collection tool. ObjectiveThe Prevention of Lower Urinary Tract Symptoms (PLUS) research consortium engaged a transdisciplinary team to build and test WhereIGo, a mobile data collection app for Android and iOS. The design goal was a comprehensive reporting system for capturing environmental, sociocultural, and physical factors that influence women’s decisions for toileting. Aims include having (1) an innovative feature for reporting physiologic urge sensation when “thinking about my bladder” and shortly before “I just peed,” (2) real-time reporting along with short look-back opportunities, and (3) ease of use anywhere. MethodsThe development team included a plain language specialist, a usability specialist, creative designers, programming experts, and PLUS scientific content experts. Both real-time and ecological momentary assessments were used to comprehensively capture influences on toileting decisions including perceived access to toileting, degree of busyness or stress or focus, beverage intake amount, urge degree, or a leakage event. The restriction on the maximal number of taps for any screen was six. PLUS consortium investigators did pilot-testing. Formal usability testing relied on the recruitment of community-dwelling women at four PLUS research sites. Women used the app for 2 consecutive days. Outcome measures were the system usability scale (SUS; 0-100 range) and the functional Mobile Application Rating Scale (1-5 range). These scales were embedded at the end of the app. The estimated a priori sample size needed, considering the SUS cut point score set at ≥74, was 40 women completing the study. ResultsFunding was provided by the National Institute of Diabetes and Digestive and Kidney Diseases since July 2015. The integrity of the build process was documented through multiple 5-minute videos presented to PLUS Consortium and through WhereIGo screenshots of the final product. Participants included 44 women, with 41 (93%) completing data collection. Participants ranged in age from 21 to 85 years, were predominantly non-Hispanic White (n=25, 57%), college-educated (n=25, 57%), and with incomes below US $75,000 (n=27, 62%). The SUS score was 78.0 (SE 1.7), which was higher than 75% of the 500 products tested by the SUS developers. The mean functional Mobile Application Rating Scale score was 4.4 (SE 0.08). The build and informal acceptability testing were completed in 2019, enrollment for formal usability testing completed by June 2020, and analysis was completed in 2022. ConclusionsWhereIGo is a novel app with good usability for women to report toileting decisions, urination, and fluid intake. Future research using the app could test the influence of real-time factors on bladder health. International Registered Report Identifier (IRRID)RR1-10.2196/54046

Germaine Tchuente Foguem, Aurelien Teguede Keleko

Introduction Advances in Artificial Intelligence (AI) offer new Information Technology (IT) opportunities in various applications and fields (industry, health, etc.). The medical informatics scientific community expends tremendous effort on the management of diseases affecting vital organs making it a complex disease (lungs, heart, brain, kidneys, pancreas, and liver). Scientific research becomes more complex when several organs are simultaneously affected, as is the case with Pulmonary Hypertension (PH), which affects both the lungs and the heart. Therefore, early detection and diagnosis of PH are essential to monitor the disease's progression and prevent associated mortality. Method The issue addressed relates to knowledge of recent developments in AI approaches applied to PH. The aim is to provide a systematic review through a quantitative analysis of the scientific production concerning PH and the analysis of the networks of this production. This bibliometric approach is based on various statistical, data mining, and data visualization methods to assess research performance using scientific publications and various indicators (e.g., direct indicators of scientific production and scientific impact). Results The main sources used to obtain citation data are the Web of Science Core Collection and Google Scholar. The results indicate a diversity of journals (e.g., IEEE Access, Computers in Biology and Medicine, Biology Signal Processing and Control, Frontiers in Cardiovascular Medicine, Sensors) at the top of publications. The most relevant affiliations are universities from United States of America (Boston Univ, Harvard Med Sch, Univ Oxford, Stanford Univ) and United Kingdom (Imperial Coll London). The most cited keywords are “Classification”, “Diagnosis”, “Disease”, “Prediction”, and “Risk”. Conclusion This bibliometric study is a crucial part of the review of the scientific literature on PH. It can be viewed as a guideline or tool that helps researchers and practitioners to understand the main scientific issues and challenges of AI modeling applied to PH. On the one hand, it makes it possible to increase the visibility of the progress made or the limits observed. Consequently, it promotes their wide dissemination. Furthermore, it offers valuable assistance in understanding the evolution of scientific AI activities applied to managing the diagnosis, treatment, and prognosis of PH. Finally, ethical considerations are described in each activity of data collection, treatment, and exploitation to preserve patients' legitimate rights.

Raghid El-Yafouri, Leslie Klieb, Valérie Sabatier

Meeting expectations is a proxy for satisfaction. With the widely used Electronic Medical Record (EMR) systems, the focus should be on whether they meet the expectations of healthcare workers and physicians. Data from a quantitative survey amongst physicians, not practices, on adopting EMR systems are analyzed with Structural Equation Modeling (SEM), K-means (K = 2) cluster analysis, and Binary Logistic Regression. The goal is to understand what factors influence meeting expectations and if these parameters are sufficient to predict the physicians’ positive or negative experiences. A path diagram shows that office-related metrics (staff skills, training, and the system’s ease of use) are the largest determinant of meeting expectations as experienced by the physicians. Two clusters centered around opposite negative and positive answers from physicians on the quality of the systems were found. The clusters provide a considerable increase in accuracy compared with the baseline in classifying positive against negative responses on meeting expectations. Physicians value the impact of EMR systems on the practice and personnel more than their own experience. We conclude medical practices should be aware of the difficulties their staff may face before implementing and during the use of EMR systems.

Sungyoung Lee, Choong-Hyun Sun, Heejun Jang et al.

Abstract Internal tandem duplication (ITD) of the FMS-like tyrosine kinase (FLT3) gene is associated with poor clinical outcomes in patients with acute myeloid leukemia. Although recent methods for detecting FLT3-ITD from next-generation sequencing (NGS) data have replaced traditional ITD detection approaches such as conventional PCR or fragment analysis, their use in the clinical field is still limited and requires further information. Here, we introduce ITDetect, an efficient FLT3-ITD detection approach that uses NGS data. Our proposed method allows for more precise detection and provides more detailed information than existing in silico methods. Further, it enables FLT3-ITD detection from exome sequencing or targeted panel sequencing data, thereby improving its clinical application. We validated the performance of ITDetect using NGS-based and experimental ITD detection methods and successfully demonstrated that ITDetect provides the highest concordance with the experimental methods. The program and data underlying this study are available in a public repository.

R. González, F.X. Aymerich, M. Alberich et al.

Background and purpose: An impaired neurovascular coupling has been described as a possible player in neurodegeneration and cognitive decline. Migraine is a recurrent and incapacitating disorder that starts early in life and has shown neurovascular coupling abnormalities. Despite its high prevalence, the physiology and underlying mechanisms are poorly understood. In this context, new biomarkers from magnetic resonance imaging (MRI) are needed to bring new knowledge into the field. The aim of this study was to determine the vein density from Susceptibility-Weighted Imaging (SWI) MRI, in subjects with migraine and healthy controls; and to assess whether it relates to Resting-State functional MRI (RS-fMRI). Materials and methods: The cohort included 30 healthy controls and 70 subjects with migraine (26 episodic, 44 chronic) who underwent a brain 3.0 T MRI. Clinical characteristics were also collected. Maps of density of veins were generated based on a Mamdani Fuzzy-Type Rule-Based System from the SWI MRI. Mean values of vein density were obtained in grey (GM) and white matter (WM) Freesurfer lobar parcellations. The Amplitude of Low-Frequency Fluctuations (ALFF) image was calculated for the RS-fMRI, and the mean values over the parcellated GM lobes were estimated. Differences between groups were assessed through and analysis of variance (age, sex, education and anxiety as covariates; p < 0.05), followed by post-hoc comparisons. Associations were run between clinical and MRI-derived variables. Results: When comparing the density of veins in GM, no differences between groups were found, neither associations with clinical variables. The density of veins was significantly higher in the WM of the occipital lobe for subjects with chronic migraine compared to controls (30%, p < 0.05). WM vein density in either frontal, temporal or cingulate regions was associated with clinical variables such as headache days, disability scores, and cognitive impairment (r between 0.25 and 0.41; p < 0.05). Mean values of ALFF did not differ significantly between controls and subjects with migraine. Strong significant associations between vein density and ALFF measures were obtained in most GM lobes for healthy subjects (r between 0.50 and 0.67; p < 0.05), instead, vein density in WM was significantly associated with ALFF for subjects with migraine (r between 0.32 and 0.58; p < 0.05). Conclusions: Results point towards an increase in vein density in subjects with migraine, when compared to healthy controls. In addition, the association between GM vein density and ALFF found in healthy subjects was lost in migraine. Taken together, these results support the idea of abnormalities in the neurovascular coupling in migraine. Quantitative SWI MRI indicators in migraine might be an interesting target that may contribute to its comprehension.

Xiaohong Wang, Xiaoli Jing, Fangkun Dou et al.

Abstract Background Biological research is generating high volumes of data distributed across various sources. The inconsistent naming of proteins and their encoding genes brings great challenges to protein data integration: proteins and their coding genes usually have multiple related names and notations, which are difficult to match absolutely; the nomenclature of genes and proteins is complex and varies from species to species; some less studied species have no nomenclature of genes and proteins; The annotation of the same protein/gene varies greatly in different databases. In summary, a comprehensive set of protein/gene synonyms is necessary for relevant studies. Results In this study, we propose an approach for protein and its encoding gene synonym integration based on protein ontology. The workflow of protein and gene synonym integration is composed of three modules: data acquisition, entity and attribute alignment, attribute integration and deduplication. Finally, the integrated synonym set of proteins and their coding genes contains over 128.59 million terminologies covering 560,275 proteins/genes and 13,781 species. As the semantic basis, the comprehensive synonym set was used to develop a data platform to provide one-stop data retrieval without considering the diversity of protein nomenclature and species. Conclusion The synonym set constructed here can serve as an important resource for biological named entity identification, text mining and information retrieval without name ambiguity, especially synonyms associated with well-defined species categories can help to study the evolutionary relationships between species at the molecular level. More importantly, the comprehensive synonyms set is the semantic basis for our subsequent studies on Protein–protein Interaction (PPI) knowledge graph.

Nicolás Gaggion, Rodrigo Echeveste, Lucas Mansilla et al.

It has recently been shown that deep learning models for anatomical segmentation in medical images can exhibit biases against certain sub-populations defined in terms of protected attributes like sex or ethnicity. In this context, auditing fairness of deep segmentation models becomes crucial. However, such audit process generally requires access to ground-truth segmentation masks for the target population, which may not always be available, especially when going from development to deployment. Here we propose a new method to anticipate model biases in biomedical image segmentation in the absence of ground-truth annotations. Our unsupervised bias discovery method leverages the reverse classification accuracy framework to estimate segmentation quality. Through numerical experiments in synthetic and realistic scenarios we show how our method is able to successfully anticipate fairness issues in the absence of ground-truth labels, constituting a novel and valuable tool in this field.

Xianjing Liu, Bo Li, Meike W. Vernooij et al.

This study addresses the challenges of confounding effects and interpretability in artificial-intelligence-based medical image analysis. Whereas existing literature often resolves confounding by removing confounder-related information from latent representations, this strategy risks affecting image reconstruction quality in generative models, thus limiting their applicability in feature visualization. To tackle this, we propose a different strategy that retains confounder-related information in latent representations while finding an alternative confounder-free representation of the image data. Our approach views the latent space of an autoencoder as a vector space, where imaging-related variables, such as the learning target (t) and confounder (c), have a vector capturing their variability. The confounding problem is addressed by searching a confounder-free vector which is orthogonal to the confounder-related vector but maximally collinear to the target-related vector. To achieve this, we introduce a novel correlation-based loss that not only performs vector searching in the latent space, but also encourages the encoder to generate latent representations linearly correlated with the variables. Subsequently, we interpret the confounder-free representation by sampling and reconstructing images along the confounder-free vector. The efficacy and flexibility of our proposed method are demonstrated across three applications, accommodating multiple confounders and utilizing diverse image modalities. Results affirm the method's effectiveness in reducing confounder influences, preventing wrong or misleading associations, and offering a unique visual interpretation for in-depth investigations by clinical and epidemiological researchers. The code is released in the following GitLab repository: https://gitlab.com/radiology/compopbio/ai_based_association_analysis}

Minz Won, Yun-Ning Hung, Duc Le

This paper investigates foundation models tailored for music informatics, a domain currently challenged by the scarcity of labeled data and generalization issues. To this end, we conduct an in-depth comparative study among various foundation model variants, examining key determinants such as model architectures, tokenization methods, temporal resolution, data, and model scalability. This research aims to bridge the existing knowledge gap by elucidating how these individual factors contribute to the success of foundation models in music informatics. Employing a careful evaluation framework, we assess the performance of these models across diverse downstream tasks in music information retrieval, with a particular focus on token-level and sequence-level classification. Our results reveal that our model demonstrates robust performance, surpassing existing models in specific key metrics. These findings contribute to the understanding of self-supervised learning in music informatics and pave the way for developing more effective and versatile foundation models in the field. A pretrained version of our model is publicly available to foster reproducibility and future research.

Hussain Ahmad Madni, Rao Muhammad Umer, Gian Luca Foresti

Federated Learning (FL) is an evolving machine learning method in which multiple clients participate in collaborative learning without sharing their data with each other and the central server. In real-world applications such as hospitals and industries, FL counters the challenges of data heterogeneity and model heterogeneity as an inevitable part of the collaborative training. More specifically, different organizations, such as hospitals, have their own private data and customized models for local training. To the best of our knowledge, the existing methods do not effectively address both problems of model heterogeneity and data heterogeneity in FL. In this paper, we exploit the data and model heterogeneity simultaneously, and propose a method, MDH-FL (Exploiting Model and Data Heterogeneity in FL) to solve such problems to enhance the efficiency of the global model in FL. We use knowledge distillation and a symmetric loss to minimize the heterogeneity and its impact on the model performance. Knowledge distillation is used to solve the problem of model heterogeneity, and symmetric loss tackles with the data and label heterogeneity. We evaluate our method on the medical datasets to conform the real-world scenario of hospitals, and compare with the existing methods. The experimental results demonstrate the superiority of the proposed approach over the other existing methods.