BackgroundHealth physical examinations play a crucial role in early detection of cancer and chronic disease. However, privacy concerns limit the utilization of this kind of data for health interventions and research. Synthetic data methods based on differential privacy are increasingly used to create complete datasets that protect privacy while enabling data analysis and result interpretation. Hence, the use of synthetic algorithms based on differential privacy for privacy protection of physical examination data is a promising research direction.MethodsThree synthetic algorithms, PrivBayes, PeGS, and DP-Gibbs were used to generate complete synthetic datasets that adhere to differential privacy standards using physical examination data composed of categorical data, which compared with the existing algorithm Private-PGM.ResultsCompared with the existing algorithm, DP-Gibbs can provide privacy preserving capacity of 4.686 (ε = 0.5), while the existing algorithm only with 2.012. In addition, DP-Gibbs provides 0.620 of precision, 0.539 of F1-score, 0.342 of Kappa Coefficient, and 0.765 of AUC-score. The corresponding statistical results of existing algorithm are 0.520, 0.321, 0.188 and 0.695.ConclusionsThe main contributions of this study are the exploration of combination models incorporating different noise forms and Bayesian synthetic algorithms, alongside a comparative analysis against existing algorithms. This study explored the balance between privacy protection and data utility under different levels of privacy protection, and DP-Gibbs offers more stable technical support for de-identifying physical examination data prior to sharing and analysis, which realized the mining and application of a wider range of medical data under the requirements of privacy protection. By leveraging this effective privacy protection technique, clinical researchers can extract valuable insights on diseases and population health from the physical examination data without the risk of leaking private information.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12911-025-03109-1.

Regional citrate anticoagulation (RCA) is critical for extracorporeal anticoagulation in continuous renal replacement therapy done at the bedside. To make patients’ data more secure and to help with computer-based monitoring of dosages, we suggest a system that uses machine learning. This system will give early alerts about citric acid overdose and advise changes to how much citrate and calcium gluconate are infused into the patient’s body. Citric acid overdose causes significant clinical risks, emphasizing the need for better adaptable anticoagulation procedures that can respond quickly. The study puts forward a new structure that uses edge computing and federated learning to make better citrate anticoagulation procedures. We proposed the resource-aware Federated Learning with Dynamic Client Selection (RAFL-Fed) algorithm in our method. In this setup, every client takes part by training a local model locally and then sending its outcome to a main server. The algorithm chooses clients for each training session depending on their computing resources, which keeps things efficient and scalable. The server collects the client inputs using weighted averages to update the global model. This step is performed repeatedly across many communication cycles, letting the system adjust to changing data trends from different locations. We put RAFL-Fed to the test on the MIMIC-IV dataset, and it outperformed other methods, getting a high accuracy of 0.9615 (IID) and 0.9571 (Non-IID), also with the lowest loss values being 0.2625 and 0.2469 in that order. It also noted the best MAE at 0.1731 (Non-IID) and a bit higher at 0.2081 (IID). Along with the high sensitivity at 0.9968, specificity stood strong as well, measuring 0.9449, plus latency was only 0.123s, which shows how effective it is for early detection of citric acid overdose as well as adjusting in real-time in the regional citrate anticoagulation process. The proposed method shows a promising solution for the real-time monitoring and adjustment of citrate anticoagulation regimens, greatly enhancing patient data security and treatment effectiveness in clinical settings. This method signifies a significant advancement in handling anticoagulation therapy.

Pregnant women need sufficient information and explanations from midwives to create a birth plan that includes pain relief options during labor. This study aimed to create an Instagram version prototype of the decision aid, “Aids of What You Can Do During Labor”, which supports the choice of various pharmacological and non-pharmacological pain relief methods. Additionally, the study evaluated the usefulness of this prototype among pregnant women planning vaginal deliveries after 28 weeks of gestation. This was a descriptive study. First, the Instagram prototype was developed on the basis of the “Aids of What You Can Do During Labor”, created by Takahashi (2023). The prototype consisted of explanations of the aid, definitions of key terms, and four steps; Step 1: “Learning how to make an informed decision”, Step 2: “Learning the pros and cons of each option”, Step 3: “Setting clear priorities for decision-making”, and Step 4: “Deciding how to spend time during labor”. A total of 17 posts containing information from Step 1 to Step 4 were created. Next, the usefulness of the prototype was assessed through a questionnaire survey conducted with 20 women with singleton pregnancies planning vaginal deliveries after 28 weeks of gestation. The usefulness data, such as whether pregnant women have found a method of labor pain relief that they would like to try were descriptively analyzed and qualitative including free-text data were analyzed via the KJ. The ethics committee approved the study. The study participants included 12 primiparas and 8 multiparas (a total of 20). More than 80% of the participants provided positive feedback regarding the usefulness of the Instagram prototype, particularly in terms of the following: “I was able to learn what I wanted to know compared to before reading the aid,” “I found a pain relief method that interested me or that I wanted to try,” and “I think reading the aid helped in creating a birth plan.” On the basis of the participants’ feedback, a revised version of the Instagram aid was created, increasing the total number of posts to 18. The usefulness of the Instagram version prototype was confirmed, through this study. Future studies should evaluate the impact of using the Instagram aid on the creation of birth plans. Not applicable.

BackgroundHeart failure (HF) represents a pressing global health issue demanding innovative and accessible approaches for early detection. Non-invasive, rapid, and cost-effective techniques utilizing deep learning (DL) hold significant promise for addressing this challenge.MethodsThis study included 462 participants categorized into healthy, coronary artery disease (CAD), and HF groups. Raw radial artery pulse wave data underwent preprocessing, including denoising, normalization, and SMOTE-based balancing. Four deep learning algorithms adept at handling sequential data – Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU), and Bidirectional Long Short-Term Memory (Bi-LSTM) – were subsequently applied and evaluated for their classification performance. A 10-fold cross-validation strategy was employed to ensure a robust evaluation of model performance and stability.ResultsTo robustly evaluate performance and stability, a 10-fold cross-validation was performed. The LSTM model yielded the highest mean accuracy (0.8595 ± 0.0522) and demonstrated strong performance across other key metrics, including a high Area Under the Curve, establishing it as the most effective model in this study. To enhance model interpretability, the SHAP (SHapley Additive exPlanations) framework was utilized to determine global feature importance and explain the final LSTM model’s predictions.ConclusionOur findings strongly suggest that an LSTM-based model analyzing radial artery pulse waves can effectively differentiate between healthy, CAD, and HF states. This approach represents a simple, non-invasive, and cost-effective methodology with significant potential as a valuable strategy to aid in the early screening and detection of HF. Further investigation and broader clinical validation are warranted to confirm the robustness and real-world applicability of this promising tool.

PurposeTo determine predictors of self-efficacy in breast cancer survivors and identify vulnerable groups.MethodsThis descriptive study was conducted between November 2023 and April 2024 at three hospitals in Türkiye and involved 430 breast cancer survivors. Data were collected through face-to-face surveys using a patient identification form and the Breast Cancer Survivor Self-Efficacy Scale. This study identified patient characteristics that indicate a tendency towards higher self-efficacy using four machine learning models; Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), XGBoost (XGB).ResultsThe mean age of participants was 50.7 ± 11.5 years. Majority of the participants (n = 425) were female. AUC values were used as ranker for the machine learning models. The ranks of the models were as follows; logistic regression model (0.715), RF (0.710), SVM (0.704), and XGBoost (0.694). Education level ranked first in the LR (0.3874), RF (0.3290), and SVM (0.1250) models, and was the second most important variable in the XGB (0.2327) model. Conversely, the cancer stage stood out in the LR (0.2466) and RF (0.1935) models, ranking third and fourth, respectively, while it ranked third in SVM (0.0683) and fourth in XGB (0.1872). Additionally, comorbidity ranked third in importance in the LR (0.2213) and RF (0.1681) models, but second in SVM (0.0705) and seventh in XGB (0.1393).ConclusionThe study demonstrated that the self-efficacy of breast cancer survivors was associated with their sociodemographic and medical characteristics. These characteristics may assist healthcare professionals in enhancing the care provided to breast cancer survivors. It is of the utmost importance to consider the aforementioned patient group as being vulnerable with regard to breast cancer survivor self-efficacy. There is a clear need for a focus on this vulnerable cohort.

BACKGROUND: Identifying predictors of in-hospital mortality in patients with ST-segment elevation myocardial infarction (STEMI) is a major concern in cardiology. The aim of this study was to identify risk factors and develop a nomogram to predict in-hospital mortality in STEMI patients. MATERIALS AND METHODS: This single-center study was a retrospective cohort analysis of all STEMI patients consecutively referred to Farshchian Hospital (Hamadan Province-Iran) from April 2021 to August 2024. Four different feature selection methods were used to select common important variables in the prediction model: Boruta, Recursive Feature Elimination (RFE), Random Forest (RF) and LASSO. The uneven distribution of the different classes 2,356 (91.7%) alive and 212 (8.3%) dead) was dealt with using the SMOTE method. After splitting the data into a training (70%) and a test (30%) dataset, a multiple logistic regression model was formulated using the significant variables identified. A nomogram predicting in-hospital mortality was then constructed and validated. RESULTS: The findings indicate that age (OR = 1.05: 95% CI: 1.03–1.06), gender (female OR = 1.71: 95% CI: 1.22–2.41), length of stay (OR = 0.89: 95% CI: 0.83–0.96), blood urea nitrogen level (OR = 1.02: 95% CI: 1.00-0.03), white blood cell count (OR = 1.07: 95% CI: 1.03–1.1), Creatinine (OR = 1.74: 95% CI: 1.36–2.22), fasting blood glucose (OR = 1.007: 95% CI: 1.005–1.009), uric acid (OR = 1.07: 95% CI: 1.02–1.12), potassium (OR = 1.2: 95% CI: 0.95–1.52) and systolic blood pressure (OR = 0.98: 95% CI: 0.97–0.99) are pivotal factor in predicting in-hospital mortality in STEMI patients. The predictive model demonstrated high accuracy (84%) and excellent discriminatory ability with an AUC of 0.91. The calibration plot demonstrated the model’s strong discriminatory performance in distinguishing between the two classes. CONCLUSIONS: The development of a nomogram for reliably predicting in-hospital mortality in STEMI patients provides clinicians with a practical visual aid for identifying high-risk patients. By enabling tailored care strategies, this tool improves therapeutic precision and ultimately leads to better clinical outcomes. CLINICAL TRIAL NUMBER: Not applicable.

Business processes in healthcare are complex and multidisciplinary, involving various professional profiles and different healthcare structures, and each medical treatment may require distinct clinical pathways. Process mining can assist in discovering trajectories, verifying compliance, and enabling an understanding of the involvement of different organizational aspects. The main goal of this study is to provide a comprehensive overview of the application of process mining in healthcare. For this, a tertiary review was conducted, gathering 18 secondary reviews that addressed different aspects, such as the objectives of process mining in healthcare, types of activities and perspectives, available resources, primary medical specialties, types of medical processes, and limitations and challenges. The study reveals that process discovery is the most common activity, while the control flow was the most used perspective. The Heuristics Miner and Fuzzy Miner algorithms were the most relevant, and oncology was the medical specialty in which process mining was most used. Process mining has proven to be an effective tool for analyzing healthcare workflows, improving understanding of clinical guidelines and protocols, and supporting decision-making. However, it is necessary to deal with noisy or missing data and establish visualization mechanisms that ensure clarity in data presentation.

The integration of electronic medical records (EMRs) with artificial intelligence (AI) is enhancing medical research, particularly in real-world evidence (RWE) studies. Extracting insights from coded medical data, such as ICD-10 codes, is essential for patient characterization. Traditional techniques, such as one-hot encoding (OHE), face limitations, particularly in managing high-dimensional data. In this study, a Bidirectional Encoder Representations from Transformers (BERT) approach is introduced to encode ICD-10 diagnostic codes, significantly improving model performance and reducing dimensionality. Data from 495,269 patients who visited the Cardiology Department at Asan Medical Center between 2000 and 2020 were used. The performance of models trained with OHE and ClinicalBERT embeddings was compared. For predicting major adverse cardiovascular events within one year following percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG), the ClinicalBERT (code-embedded) model outperformed OHE. It achieved an AUC of 0.746 compared to 0.719, while also significantly reducing the dimensionality from 2,492 to 128. This method, which integrates diagnostic and medication data, provides valuable insights into patient care, enhancing the precision of predictions and supporting healthcare professionals in making more informed decisions.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12911-025-03145-x.

BackgroundChronic wounds (CWs) represent a significant and growing challenge in healthcare due to their prolonged healing times, complex management, and associated costs. Inadequate wound assessment by healthcare professionals (HCPs), often due to limited training and high clinical workload, contributes to suboptimal treatment and increased risk of complications. This study aimed to develop an artificial intelligence (AI)-powered wound assessment tool, integrated into a mobile application, to support HCPs in diagnosis, monitoring, and clinical decision-making.MethodsA multicenter observational study was conducted across three healthcare institutions in Western Switzerland. Researchers compiled a hybrid dataset of approximately 4,000 wound images through both retrospective extraction from clinical records and prospective collection using a standardized mobile application. The prospective data included high-resolution images, short videos, and 3D scans, along with structured clinical metadata. Retrospective data were anonymized and manually annotated by wound care experts. All images were labeled for wound segmentation and tissue classification to train and validate deep learning models.ResultsThe resulting dataset represented a broad spectrum of wound types (acute and chronic), anatomical locations, skin tones, and healing stages. The AI-based wound segmentation model, developed using the Deeplabv3 + architecture with a ResNet50 backbone, achieved a DICE score of 92% and an Intersection-over-Union (IOU) score of 85%. Tissue classification yielded a preliminary mean DICE score of 78%, although accuracy varied across tissue types, especially fibrin and necrosis. The models were optimized for mobile implementation through quantization, achieving real-time inference with an average processing time of 0.3 seconds and only a 0.3% performance reduction. The dual approach to data collection, prospective and retrospective—ensured both image standardization and real-world variability, enhancing the model’s generalizability.ConclusionsThis study laid the foundation for an AI-driven digital tool to assist clinical wound assessment and education. The integration of robust datasets and AI models demonstrated the potential to improve diagnostic precision, support personalized care, and reduce wound-related healthcare costs. Although challenges remained, particularly in tissue classification, this work highlighted the promise of AI in transforming wound care and advancing clinical training.Trial registrationNot applicable.