IntroductionThe analysis of electrocardiograms (ECGs) using deep learning has made significant progress in recent years, whereby the variability in ECG data due to different recording techniques introduces a significant challenge. ECGs can be recorded in various modalities, including telemedical (tECG), resting (rECG), and long-term (hECG) ECGs, each with unique characteristics and applications. tECGs, commonly used in remote patient monitoring for conditions such as heart failure, are recorded by patients at home and use a limited number of non-standardized leads. rECGs are typically short recordings taken in clinical settings using standardized leads, while hECGs involve continuous recording over 24 hours or more. The resulting differences in data quality, duration, and clinical context can significantly impact model performance.PurposeWe aim to develop cross-modal models with robust indicators of failure points when inferring on modalities underrepresented during training. We hypothesize that modality imbalance during training significantly affects the generalization ability of deep learning models, with model confidence as an indicator of domain shift.MethodsWe develop a deep learning model by training and evaluating on ECG data from three distinct modalities. Our approach involves using publicly available datasets for rECGs and proprietary datasets for tECGs and hECGs. We systematically vary the ratio of training data from each modality and analyze its effect on model performance and generalization. To assess model reliability under domain shift, we apply Monte Carlo (MC) Dropout to estimate predictive uncertainty. This allows us to quantify the confidence of the model across modalities.ResultsOur experiments show significant differences in model performance depending on the ECG modality we use for training and evaluation. When we train models predominantly on data from a single ECG modality, they often show reduced performance when applied to another, highlighting challenges in cross-modality generalization. For example, balancing the training data distribution further toward tECGs improved accuracy on a telemedical dataset (500 sinus rhythm and 500 atrial fibrillation ECGs) from 91.7% to 94.8%. MC Dropout consistently estimates increased model uncertainty associated with this domain shift.ConclusionOur study highlights that ECG modality is not only a technical detail but a key factor influencing the performance and reliability of deep learning based ECG analysis. The modality has a substantial impact on the model performance and generalizability. To address this, we propose an approach that shows high robustness across diverse ECG types. Our method recognizes modality-specific characteristics within ECG recordings and integrates uncertainty estimation to improve the performance stability and reliability of deep learning models in real-world ECG analysis.

BackgroundChildhood cancer is frequently treated with anthracycline chemotherapies and chest-directed radiation. While such treatments have improved five-year survival rates, they are also cardiotoxic, placing survivors at lifelong risk for cardiovascular diseases such as cardiomyopathy and heart failure. Recent literature suggests that analysis of electrocardiograms (ECG) using artificial intelligence (AI) methods, also known as ECG-AI, applied to lead I of a standard clinical ECG can predict levels of brain natriuretic peptides.PurposeWe hypothesize and test that an existing ECG-AI model can also detect cancer survivors with elevated NT-proBNP levels using a single-lead Apple Watch ECG as the sole model input.MethodsWe previously developed an ECG-AI model capable of estimating BNP from lead I of a standard 12-lead clinical ECG using a large dataset of same-day ECG-BNP pairs from Wake Forest School of Medicine, Winston-Salem, NC. Also, as part of an ongoing study (Akbilgic & Hudson), we collected paired same-day single-lead Apple Watch ECGs and NT-proBNP from participants in the St. Jude Lifetime Cohort (SJLIFE), a clinically assessed cohort of adult survivors of childhood cancer diagnosed and treated at St. Jude Children’s Research Hospital, Memphis, TN between 1962-2012. Without any fine-tuning, we applied the ECG-AI model to the Apple Watch ECGs and calculated the area under the receiver operating characteristic curve (AUC) and other accuracy metrics for the binary outcome of elevated (>300 pg/mL) versus non-elevated (≤300 pg/mL) NT-proBNP.ResultsOur analytical cohort included Apple Watch–NT-proBNP pairs from 580 SJLIFE participants: 82.2% White, 13.3% Black, and 49.5% female, with a mean age ± standard deviation of 37 ± 10 years. Among them, 16 participants (2.8%) had NT-proBNP levels >300 pg/mL. Applying the ECG-AI model for BNP detection achieved an AUC of 0.84 (95% CI: 0.75–0.95) for identifying elevated NT-proBNP levels (Figure 1).ConclusionsThe ECG-AI model—although not specifically designed for cancer survivors, wearable ECG devices, or NT-proBNP detection—was able to identify survivors with elevated NT-proBNP levels with high accuracy. Future work will focus on developing a model specifically tailored for detecting elevated NT-proBNP among at-risk cancer survivors, with the goal of enabling low-cost, non-invasive, and improved lifelong cardiovascular surveillance.Elevated NT-proBNP Detection Accuracies

IntroductionThe emergence of wearable Electrocardiogram (ECG) recorders enables real-time analysis and detection of arrhythmias. Deep Learning (DL) techniques can be used to analyze the large volumes of ECG data produced by wearable devices. However, the application of DL techniques in real-world clinical environments is challenging due to limited access to raw ECG data and the rapid change in wearable hardware. DL techniques require annotated training data on the new hardware to maintain performance, which is costly to obtain. To address these challenges, we propose a visual algorithm for cross-dataset ECG classification using metric learning techniques and an extended ECG context window to improve performance without the need for signal filtering.MethodsA combination of three public datasets is utilized for training, which are MIT-BIH Arrhythmia, MIT-BIH Atrial Fibrillation, and A Large-scale 12-lead arrhythmia database, producing a training dataset containing 124,080 ECG segments. The segments are 10-seconds in duration and comprise four arrhythmia categories: Normal, Supraventricular, Ventricular and Atrial Fibrillation. The training data is denoised using signal filtering and horizontal flipping is introduced to improve the model's generalizability. The model is tested using the Long-term Atrial Fibrillation Database, without signal filtering to reflect real-world conditions. A two-stage metric learning algorithm is proposed whereby ECG features are extracted using a residual Siamese network and the extended context window classification algorithm with a 30-second ECG data window is used to classify the ECG segments.ResultsThe proposed extended context window algorithm with horizontal flipping augmentation achieved an accuracy of 88.28% and a Macro-F1 score of 78.30%. Compared to the baseline model that achieved an accuracy of 85.70% and a Macro-F1 score of 71.97%, the proposed extended context algorithm yields superior performance and serves as a replacement for signal denoising.ConclusionThe proposed algorithm for extending the ECG context window proves to be an effective solution for improving DL algorithm performance in the absence of signal filtering, a common occurrence in real-world clinical environments such as the analysis of document-based ECG reports.Overview of the training approachArchitecture of the proposed algorithm

Multi-scale heart simulation augments the explainability of artificial intelligence-enabled electrocardiogram through provision of an electrocardiogram database labelled with cellular pathologies.

ECG-aBcDe: Overcoming model dependence, encoding ECG into a universal language for any large language model.

BackgroundCardiovascular diseases remain the leading global cause of mortality, yet traditional electrocardiogram (ECG) interpretation shows subjective variability and limited sensitivity to complex pathologies.ObjectiveThis study aims to address these challenges by proposing the Cardiovascular Multimodal Prediction Network (CaMPNet), a transformer-based multimodal architecture that integrates raw 12-lead ECG waveforms, 9-structured machine-measured ECG features, and demographic data (age and sex) through cross-attention fusion.MethodsThe model was trained on 384,877 records from the Medical Information Mart for Intensive Care IV - Electrocardiogram Matched Subset database and evaluated across 12 cardiovascular disease labels. To further assess temporal robustness, a temporal external validation was performed using the most recent 10% of the data, withheld chronologically from model development.ResultsOn the internal test set, the model achieved a mean area under the curve (AUC) of 0.845 (SD 0.04) and area under the precision-recall curve of 0.489, outperforming the residual networks-ECG baseline (AUC=0.848 but F1-score=0.152) and all single-modality variants. Subgroup analyses demonstrated consistent performance across demographics (male AUC= 0.846 vs female=0.843; youngest quartile 0.884 vs oldest 0.811). CaMPNet retained moderate discriminative ability in temporal external validation with a mean AUC of 0.715 (SD 0.03) and area under the precision-recall curve of 0.298, although performance declined due to temporal distribution shifts. Despite this, major disease categories, such as atrial fibrillation, heart failure, and normal rhythm, maintained high AUCs (>0.84). Attention-based visualization revealed clinically interpretable patterns (eg, ST-segment elevations in ST-segment elevation myocardial infarction), and ablation experiments verified the model’s tolerance to missing structured inputs.ConclusionsCaMPNet demonstrates robust and interpretable multimodal ECG-based diagnosis, offering a scalable framework for comorbidity screening and continual learning under real-world temporal dynamics.

Automated diagnosis of chronic obstructive pulmonary disease using deep learning applied to electrocardiograms

Cardiovascular diseases (CVDs) remain a leading cause of mortality worldwide, necessitating the identification and analysis of key risk factors. This study examines the epidemiological effects of cardiovascular risk factors, such as heart rate variability (HRV) and electrocardiogram (ECG) analysis, diabetes, smoking, alcohol use, obesity, and hypertension focusing on cardiovascular diseases (CVD) in India, with a specific emphasis on the state of Madhya Pradesh. It investigates the relationship between heart rate (HR) and various risk factors associated with cardiovascular problems, along with an assessment of heart rate variability (HRV) as an indicator of cardiovascular health. Through a comprehensive analysis of epidemiological data, the study identifies key risk factors contributing to the prevalence of CVD in the region, including lifestyle habits, socio-economic factors, and genetic predispositions. Utilizing both crosssectional and longitudinal data, the research evaluates the impact of these risk factors on HR and HRV. Findings highlight significant correlations between elevated HR, reduced HRV, and increased cardiovascular risk, providing insights into the potential for HRV assessment as a diagnostic tool for early detection and management of CVD. This study aims to contribute to a deeper understanding of cardiovascular health dynamics in Madhya Pradesh and offer evidencebased recommendations for public health interventions.

Introduction. Every year, more than 1 million limb amputations are recorded worldwide, with the main causes being vascular diseases and injuries. Since the beginning of the large-scale war the number of amputees in Ukraine has rapidly increased. They require long-term rehabilitation and monitoring of vital functions. Aim. To investigate the peculiarities of electrocardiogram (ECG) recording and analysis in military personnel with limb amputations during rehabilitation. Materials and methods. The ECGs of 48 military personnel with various levels of limb amputations in inpatient rehabilitation were analyzed; 5 cases are presented. Results. The ECGs of five military personnel with different levels of limb amputations were analyzed. When assessing the voltage of the ECG complexes and analyzing their waves, intervals and segments, changes in standard and augmented leads are noted. Differences in cardiac activity indicators were found with incorrect electrode placement, which can lead to erroneous ECG analysis due to improper recording technique during amputations. When assessing the shape, duration and amplitude of the ECG elements, an increase in the amplitude of the P wave and a decrease in the duration of the QRS complex were found, which can be regarded as gradual benign physiological changes at the initial stage after amputation. Myocardial repolarization parameters were analyzed, including the duration of the QT, QTc, Tr-e intervals, the Tr-e/QT, and Tr-e/QTc ratios, which should be considered as markers for predicting ventricular arrhythmias and cardiovascular mortality. Conclusions. During rehabilitation and active use of the prosthesis, physical exertion increases, which requires monitoring of vital functions. The prosthetics process requires rehabilitation at all stages (pre-prosthetic, interprosthetic, and prosthetic) both in inpatient and outpatient settings. Monitoring tolerance to physical exertion, which increases during rehabilitation, and the functions of vital organs is a necessary diagnostic procedure. ECG is a mandatory method of instrumental diagnostics at the rehabilitation stage, which requires its correct performance and analysis in patients with limb amputations depending on the level of amputation.

The study aimed to evaluate the diagnostic performance of artificial intelligence (AI) in detecting sudden cardiac death on electrocardiogram (ECG). We systematically searched PubMed, Web of Science, Embase, and IEEE Xplore for studies published through April 2025 evaluating AI models for ECG-based sudden cardiac death detection, using expert consensus or database records as the reference standard. A bivariate random-effects model generated pooled sensitivity and specificity estimates. Heterogeneity was quantified via I2 and τ2 statistics. Study quality was appraised using the revised QUADAS-2 tool, with evidence certainty graded via the GRADE assessment. Out of 958 initially identified studies, 27 studies with 2613 patients and images were ultimately included for the final analysis. For heart rate variability, AI demonstrated a sensitivity of 0.90 (95% CI: 0.86-0.92) and specificity of 0.91 (95% CI: 0.83-0.96), with an AUC of 0.93 (95% CI: 0.91-0.95). For ECG signal segmentation, AI demonstrated a sensitivity of 0.96 (95% CI: 0.92-0.98) and specificity of 0.99 (95% CI: 0.94-1.00), with an AUC of 0.99 (95% CI: 0.98-1.00). For direct input of ECG lead signals, AI demonstrated a sensitivity of 0.87 (95% CI: 0.61-0.97) and specificity of 0.91 (95% CI: 0.75-0.97), with an AUC of 0.95 (95% CI: 0.93-0.97). This meta-analysis indicates that AI-based ECG analysis shows potential for SCD prediction. However, the summary estimates are derived from highly heterogeneous studies and should not be considered benchmarks for clinical performance. The current evidence remains preliminary and derived from idealized research settings, underscoring the need for prospective, multicenter studies with standardized methodologies to establish generalizability and clinical applicability.

We propose an efficient deep learning model to improve the classification accuracy in automatic classification tasks of 12-lead electrocardiogram (ECG) signals. We designed a new ResLSTM-TemporalSE network architecture by incorporating a multi-layer Residual Long Short-Term Memory (ResLSTM) structure and introducing skip connections between LSTM layers to establish residual learning pathways for the temporal features. A temporal attention mechanism was integrated into the traditional Squeeze-and-Excitation (SE) module to enhance channel-wise feature representation while capturing long-term temporal dependencies within ECG signals, thereby an efficient hierarchical feature extraction framework was constructed. The model was validated using the public CPSC2018 dataset and a private clinical dataset from the Seventh Affiliated Hospital of Southern Medical University. The experimental results demonstrated that the model achieved a classification accuracy of 99.70% on the CPSC2018 test set, with precision, recall, and F1-score values of 0.9966, 0.9370, and 0.9653, respectively. On the private clinical dataset, it attained an accuracy of 82.77%, with precision, recall, and F1-score values of 0.6811, 0.8961, and 0.7723. Ablation studies confirmed the significant contributions of both the residual connections and the temporal attention module to model performance. The ResLSTM-TemporalSE model effectively integrates spatiotemporal features of the ECG signals and demonstrates superior classification performance on the CPSC2018 benchmark while maintaining strong generalization capabilities in real-world clinical settings. This framework provides a robust solution for automated ECG analysis and holds significant promise for clinical applications.

Cardiovascular diseases remain the leading cause of mortality, and automated electrocardiogram (ECG) analysis can ease clinical workloads but is limited by scarce and imbalanced data. Synthetic ECG can mitigate these issues, and while most methods use Generative Adversarial Networks (GANs), recent work show variational autoencoders (VAEs) perform comparably. We introduce cNVAE-ECG, a conditional Nouveau VAE (NVAE) that generates high-resolution, 12-lead, 10-second ECGs with multiple pathologies. Leveraging a compact channel-generation scheme and class embeddings for multi-label conditioning, cNVAE-ECG improves downstream binary and multi-label classification, achieving up to a 2% AUROC gain in transfer learning over GAN-based models.

Understanding and categorizing fetal health is an influential field of research that profoundly impacts the well-being of both mother and child. The primary desire to precisely examine and cure fetal disorders during pregnancy to enhance fetal and maternal outcomes is the driving force behind the classification of fetal health. Fetal cardiac abnormalities (structural or functional) need immediate doctor attention, and their early identification and detection in all stages of pregnancy can help doctors with the timely treatment of the mother and the unborn child by enabling appropriate prenatal counseling and management. By knowing about fetal health and taking necessary precautions for fetal health, the rate of fetal mortality can be decreased. Advancements in machine learning (ML) algorithms have revolutionized the analysis of fetal electrocardiogram (ECG) signals. MachineLearning and Deep Learning algorithms automate the fetal monitoring processanddecisions in emergencies, save time, and enable telemonitoring. This paper introduces a new hybrid approach to enhance fetal health classification using an intelligent and dynamic combination of Random Forest (RF) and AdaBoost machine learning algorithms. The proposed work includes a detailed review of existing models and the challenges in handling fetal health data, setting the foundation for the design of advanced hybrid models. The implemented algorithm effectively integrates the strengths of RF and AdaBoost to enhance fetal health monitoring and classification performance. The RF algorithm is widely established for its capacity to manage large and highly dimensional data sets, whereas AdaBoost focuses on enhancing classification accuracy by correcting for mistakes in the RF models' predictions. The proposed hybrid model is tested on a recognized benchmark CTG dataset, where it attained a classification accuracy of 95.98%, a precision of 92.88%, a recall of 92.78% and an F1 score of 92.70%. Achieved results demonstrate the potential of our novel approach in real-world applications, offering a promising tool for early detection of fetal anomalies, which is crucial for both fetal and maternal health. Fetal health classification and timely prediction of fetal diseases seem to be a critical step throughout pregnancy. So, to deal with this problem, an attempt has been made to propose an accurate, reliable, and novel hybrid approach for enhancing fetal health classification. By combining the strengths of two algorithms, named RF and AdaBoost, superior classification accuracy, precision, F1 score, and recall have been achieved, and much better robustness compared to standalone models. We have strived to make a noteworthy impact on the health sector by developing this hybrid model for the timely evaluation and prediction of fetal-maternal health.

Abstract This review presents a comprehensive examination of advanced artificial intelligence (AI) techniques for wearable electrocardiogram (ECG) analysis, with a focus on secure, generalizable and personalized solutions enabled by tiny machine learning (TinyML). As cardiovascular diseases continue to impose a significant global health burden, accurate and real-time ECG monitoring is essential. We detail how traditional and deep learning methods, ranging from convolutional neural networks (CNNs) to bio-inspired spiking neural networks (SNNs), are adapted for resource-constrained devices. Emphasis is placed on strategies to enhance model generalizability, such as leveraging broader datasets, lead reduction, improved loss functions and generative models, alongside approaches for individual personalization through fine-tuning and meta-learning. Furthermore, the review addresses critical security and privacy challenges inherent to on-device processing and federated learning, highlighting methods like homomorphic encryption, secure multi-party computation and differential privacy. By examining what seemed like separate fields, this article highlights developments toward robust, energy-efficient and privacy-aware wearable ECG monitoring systems, supporting improved diagnostics and personalized healthcare in edge-computing environments. Recently, μ-trainers, a novel technique addressing this intertwined area, has been proposed for on-device personalization: a compact trainer fine-tunes directly on personal data within the device, updating the full model for individualized inferences without transmitting sensitive information, thereby preserving privacy.

This study proposes a novel normalization method that retains signal-specific amplitude characteristics, improving the differentiation between healthy and unhealthy electrocardiogram (ECG) signals. We present a nonlinear network-based approach to analyze ECG dynamics using a recurrence network (RN), described by characteristic path length, link density, and the clustering coefficient as key network measures. To quantify signal complexity, we introduce weighted Shannon entropy measures based on the distributions of shortest path lengths and clustering coefficients in the RN. Degree heterogeneity is further investigated to examine network-level local node variability between cardiac conditions. A comprehensive analysis using the Physikalisch-Technische Bundesanstalt (PTB) Diagnostic ECG Database of Physionet shows that the proposed approach effectively differentiates between normal and abnormal signals, including bundle branch block, myocardial infarction, dysrhythmias, hypertrophy, and cardiomyopathy even when only the V1 lead is used. The method achieves a classification accuracy of 93.5% for Random Forest and 91.7% with XGBoost, confirming the robustness of recurrence-based features in short-duration ECG analysis. The recurrence-based topological and complexity measures, integrated with the proposed amplitude-preserving normalization, quantitatively agree with physiological mechanisms such as conduction delay, repolarization instability, and morphological irregularity, offering a reliable framework for real-time and clinically useful cardiac diagnosis.

This study evaluated the performance of ChatGPT 4 and 3.5 versions in answering nephrology questions from medical residency exams in Brazil. A total of 411 multiple-choice questions, with and without images, were analyzed, organized into four main themes: chronic kidney disease (CKD), hydroelectrolytic and acid-base disorders (HABD), tubulointerstitial diseases (TID), and glomerular diseases (GD). Questions with images were answered only by ChatGPT-4. Statistical analysis was performed using the chi-square test. ChatGPT-4 achieved an overall accuracy of 79.80%, while ChatGPT-3.5 achieved 56.29%, with a statistically significant difference (p < 0.001). In the main themes, ChatGPT-4 performed better in HABD (79.11% vs. 55.17%), TID (88.23% vs. 52.23%), CKD (75.51% vs. 61.95%), and DG (79.31% vs. 55.29%), all with p < 0.001. ChatGPT-4 presented an accuracy of 81.49% in questions without images and 54.54% in questions with images, with an accuracy of 60% for electrocardiogram analysis. This study is limited by the small number of image-based questions and the use of outdated examination items, reducing its ability to assess visual diagnostic skills and current clinical relevance. Furthermore, addressing only 4 areas of Nephrology may not fully represent the breadth of nephrology practice. ChatGPT-3.5 was found to have limitations in nephrology reasoning compared to ChatGPT-4, evidencing gaps in knowledge. The study suggests that further exploration is needed in other nephrology themes to improve the use of these AI tools.

Whether artificial intelligence (AI) analysis of single-lead ECG (1 L ECG) can predict incident AF is unknown. In the VITAL-AF trial (ClinicalTrials.gov NCT03515057, registered 2/24/2021) of primary care patients aged ≥65 years undergoing handheld 1 L ECG screening, we tested three AI approaches to incident AF prediction, and compared the best model to the CHARGE-AF risk score. In a test set of 4,221 individuals, a published AI model trained using single standard ECG leads (“1 L ECG-AI”) provided similar 2-year AF discrimination to models trained with VITAL-AF data. In the full VITAL-AF sample of 15,694 individuals without prevalent AF (2-year incident AF 3.1%), 1 L ECG-AI with age/sex (1 L ECG-AI AS) had comparable discrimination (area under the receiver operating characteristic curve [AUROC] 0.695[0.637–0.742]; average precision [AP] 0.060[0.050–0.078]) to CHARGE-AF (AUROC 0.679[0.623-0.730]; AP 0.062[0.052–0.080], AUROC p = 0.46, AP p = 0.92). Net reclassification improvement was favorable versus age ≥65 years (0.27[0.22–0.32]). 1 L ECG-AI may increase efficiency and reach of AF screening.

The aim – to objectivize the analysis of quantitative electrocardiogram (ECG) indicesconsidering factors such as age, sex, lipid metabolism parameters, and the effect ofdifferent levels of uric acid.Material and methods. A total of 19 patients with chronic coronary syndrome (CCS)(previously confirmed diagnosis of stable angina of functional class (FC) intensity II–III) with further distribution into groups according to age, sex, blood test parameters(levels of uric acid (UA), total cholesterol (TCL), high-density lipoprotein cholesterol(CL LPHD) and low-density lipoprotein cholesterol (CL LPLD)), echocardiographicindices (left ventricular ejection fraction (LVEF) were included in the study.Electrocardiographic assessment was performed using a graphic analysis concept thatcomprised 19 parameters of waves and intervals. These included the area andamplitude of the P wave, the minimum, maximum, and mean P-wave duration acrossall leads, P-wave dispersion, the PQ interval (from the onset of the P wave to thebeginning of the QRS complex), amplitudes of the R and S waves, QRS complexduration, T-wave amplitude, Tp–Te interval, minimum and maximum T-waveamplitudes, T-wave dispersion, as well as corrected QT interval (QTc) parameters –minimum, maximum, mean values, and QTc dispersion (Fig. 1).Statistical analysis was carried out using spreadsheet software, applying bothparametric and non-parametric methods. In the case of normally distributed data,Student’s t-test was used, with results presented as mean value (M) ± standard error ofthe mean (m). For non-normally distributed data, the Mann–Whitney U test wasapplied. Correlation analysis was performed using Pearson’s and Spearman’scorrelation coefficients. Odds ratios (OR) and relative risks (RR) were also calculated.The critical level of statistical significance was set at p &lt; 0.05, and a trend wasconsidered at p &lt; 0.1.The work was conducted in compliance with the main provisions of GCP (1996), theCouncil of Europe Convention on Human Rights and Biomedicine (1997), theDeclaration of Helsinki of the World Medical Association on the ethical principles ofmedical research involving human subjects (1964–2008), and the Order of the Ministryof Health of Ukraine No. 690 dated 23.09.2009 (with amendments introduced by OrderNo. 523 dated 12.07.2012).This work was carried out within the framework of the department’s initiative researchproject “Precision approach to the treatment of acute and chronic coronary syndromes– modern management and therapeutic perspectives” (2025–2029).Results. The study found statistically significant differences in the quantitative ECGindices depending on uric acid levels (UA, distribution ≥443&lt; mmol/L). Whenanalyzing two groups based on UA levels (368,1±18,97 mmol/L vs. 510,6±7,77mmol/L, p&lt;0,001), a significant decrease in the R wave amplitude was found withhigher UA levels (10,89±1,00 mm in the lower UA group vs. 7,1±0,72 mm in the higherUA group, p&lt;0,02, Δ%1-2=-35,00%, when determining by 100% of R wave amplitudein the lower UA group). Gender and age distributions showed a correlation between Twave amplitude and low-density lipoprotein cholesterol levels. The impact of leftventricular ejection fraction (LVEF, distribution ≥56&lt;%, which amounted to62,0±1,13% vs. 50,11±1,31%, p&lt;0,001) on ECG indices showed significantdifferences in the QRS complex duration with an increase in QRS duration at lower LVEF values (155,6±14,05 ms vs. 108,9±11,11 ms, p&lt;0,02, Δ%1-2=-30%, whendetermining by 100% of QRS duration in the lower LVEF group). Analysis based ontotal cholesterol levels (TC, distribution ≥5,46&lt; mmol/L, 6,36±0,12 mmol/L vs.3,91±0,22 mmol/L, p&lt;0,001) revealed significant differences in T wave dispersion witha decrease in the TC increase group (80,0±14,96 ms vs. 43,33±4,66 ms, p&lt;0.05, Δ%1-2=45,8%, when determining by 100% in the lower TC group).Conclusion. The key role in the identified differences in the quantitative evaluation ofECG was played by the indices of T and R wave amplitudes, T wave dispersion, andQRS complex duration depending on age, gender, lipid metabolism parameters, andUA levels.

Acute care settings, including emergency medicine and intensive care units, comprise a substantial portion of healthcare and are essential in the prompt management of conditions that can prove fatal. Critical care conditions require timely management that can be delayed by high patient volumes and the need for complex clinical decision making. Artificial intelligence (AI) tools have been created to enhance diagnostic accuracy and optimize workflow to improve patient care. This narrative review discusses the current status of AI in acute care, with a focus on its applications in triaging and diagnosis. AI-enhanced electrocardiogram analysis, identification of myocardial infarction and acute coronary syndrome, and heart failure risk stratification led to better patient-specific management and improved results. AI models successfully determined and aided in the timely management of various acute conditions, including pneumonia, pulmonary embolism, and respiratory failure. The AI algorithms used accurately determined sepsis onset and course, superseding traditionally used clinical tools and leading to early diagnosis and reduced sepsis mortality. These models showed high sensitivity and specificity in diagnosing and triaging neurological conditions, including altered levels of consciousness, seizures, and intracranial hemorrhages. AI that involved advanced machine learning imaging software led to faster and more accurate stroke diagnosis. Diagnostic tools assisted by AI improved the detection and classification of acute pancreatitis, appendicitis, and gastrointestinal bleeding. AI has shown promising results in optimizing management in acute care settings. However, critical issues in data standardization, ethical considerations, and clinical workflow integration need to be addressed to enable clinical implementation.

Prolonged P-wave duration on electrocardiogram (ECG) quantitatively reflects atrial remodeling and is associated with atrial fibrillation, cardiovascular death. We examined the association between P-wave duration and the risk of cardiovascular events with and without left ventricular hypertrophy (LVH) on ECG. The COUPLING Study, which examined prognosis in a Japanese population at cardiovascular risk, included 4288 subjects for whom digital ECG analysis was available. The primary event was a composite endpoint of stroke, ischemic heart disease, sudden death, hospitalization for heart failure, and aortic dissection. The cutoff value of the P-wave in ECG was 140 ms, and patients were divided into two groups: a control group with normal P-wave duration (n = 3975) and a prolonged P-wave duration group (n = 313). The association between prolonged P-wave duration and cardiovascular events was investigated in the presence of LVH using the Cornell product criteria for ECG. The mean age of subjects was 69 ± 11 years, and 50% were male. The median follow-up period was 5.0 years, and primary events were observed in 178 patients. The hazard ratio of primary events for prolonged P-wave duration was 2.20 (95% confidence interval [CI] 1.47-3.29, p < 0.001) after adjusting for age, gender, and comorbidity. In patients without LVH, prolonged P-wave duration was associated with a 1.86-fold higher primary endpoint risk (95% CI 1.17-2.96, p = 0.008), while prolonged P-wave duration was associated with a 2.76-fold higher primary event risk in patients with LVH (95% CI 1.10-6.89, p = 0.030), and the association was stronger in patients with LVH on ECG (synergistic effect: p = 0.007). Cardiovascular risk factors can advance atherosclerosis. Elevated sympathetic nerve activity affects the development of atrial fibrillation, and the renin-angiotensin-aldosterone system affects atrial and ventricular filling pressures as well as volume overload. In these processes, prolonged P-wave, reflecting atrial remodeling, and BNP, reflecting impaired ventricular function, indicates that atherosclerosis is developing. Prolonged P-wave in advanced ECG-LVH is a risk for the occurrence of cardiovascular events.