European Heart Journal - Digital Health

Background Artificial intelligence-enhanced electrocardiography (AI-ECG) enables scalable, low-cost cardiac dysfunction screening, but existing models are annotation-intensive and predominantly adult-derived, leaving paediatric generalizability uncertain. Paediatric cohorts exhibit highly variable cardiac morphology and function compared to adults, which may be useful for learning generalizable AI-ECG models. Methods We pretrained ECG-Fyler on a predominantly paediatric, all-age cohort at Boston Children's Hospital (1992-2023), annotated with a cardiology-specific coding system (Fyler codes), and evaluated it on assessments from echocardiography (echo) and cardiac magnetic resonance (CMR) studies. We validated on an external adult cohort from Columbia University Irving Medical Center. Performance was benchmarked against several AI-ECG foundation models by AUROC across age groups, lesion types, and limited-data scenarios. Findings The pretraining cohort comprised 782,138 ECGs from 255,271 patients (median age: 10.9 years, IQR: [2.8-16.8]). Internal evaluation included 178,495 ECG-echo pairs (median age: 10.9 [3.7-17.0]) and 8,584 ECG-CMR pairs (median age: 20.7 [15.6-29.6]). External validation included 82,543 ECG-echo pairs from adults (median age: 64.0 [52.0-74.0]). ECG-Fyler improved AUROC across biventricular dysfunction and dilation tasks, with the largest gains in low-data settings. In internal validation, ECG-Fyler detected low left ventricular ejection fraction (LVEF [&le;] 40%) from only 100 fine-tuning samples (AUROC: 0.80, 95% CI: [0.78-0.80]), outperforming other models (AUROC < 0.65) and improving with additional fine-tuning (AUROC: 0.94 [0.93-0.94]). Similar improvements were observed for CMR-derived LVEF, RVEF, and ventricular dilation. In external validation on adults, ECG-Fyler exhibited an AUROC of 0.83 (CI: [0.82-0.85]) for LVEF [&le;] 40%. After fine-tuning on less than 10% of external data, LVEF [&le;] 45% performance (AUROC: 0.87 [0.86-0.88]) outperformed a fully trained, site-specific prior model (AUROC: 0.85 [0.84-0.87]). Interpretation Pretraining on richly annotated, paediatric-dominant ECGs yields models that transfer efficiently across institutions and ages, supporting AI-ECG screening and triage when labels or imaging access are limited. Funding National Institutes of Health (R01LM012973); Kostin Innovation Fund, Boston Children's Hospital

Cardiovascular disease is the leading cause of death worldwide. Sudden cardiac death (SCD) accounts for roughly 50% of all cardiac deaths. The electrocardiogram (ECG) is widely used for early diagnosis of cardiac disease. However, the complexity of accurate interpretation limits the ECG's efficacy. Modern deep learning methods have been applied to assist clinicians in diagnosis. We applied Neural Architecture Search (NAS), an automated machine learning technique, to identify optimal deep learning architectures for classifying cardiac arrhythmias from ECGs. We applied the Differentiable Architecture Search strategy to an AutoFormer search space to identify optimal self-attention architectures for arrhythmia classification. We trained, validated, and tested the resulting model on the PhysioNet Challenge 2021 dataset (n = 88,253), comprising ECGs across three continents. We performed a hyperparameter optimisation on the NAS output, exploring input patch size, class weighting, and loss function. We evaluated performance using the PhysioNet Challenge metric and the area under the receiver operating characteristic curve (AUROC). The NAS converged towards minimal architectural configurations (embedding dimension: 384, depth: 4, self-attention heads: 4, MLP ratio: 1) with a validation challenge metric of 0.66 (PhysioNet Challenge 21 Winner: 0.63). The NAS-created network achieved an AUROC of 0.97 and a challenge metric of 0.71 during testing. Normal Sinus Rhythm and Sinus Tachycardia achieved AUROCs of 0.99. Low-QRS Voltage and T-wave abnormality were the worst-performing arrhythmias, with AUROCs of 0.89 and 0.90, respectively. We interpret that architectural simplicity drives performance in arrhythmia classification. Because SCD is unexpected, prevention strategies in free-living environments require lightweight computational resources suitable for wearable devices. Class imbalance fundamentally limits classification performance for rare arrhythmias such as Low-QRS Voltage and T-wave inversion, irrespective of hyperparameter choices. However, the self-attention mechanism can autonomously abstract clinical representations, simplifying clinical deployment by eliminating the need for an explicit feature-extraction pipeline.

Diastolic heart failure (HF) in primary cardiomyopathy is under-recognized and often diagnosed late, particularly in children. While recent studies have advanced understanding of HF with preserved ejection fraction in older adults, the prevalence, outcomes and molecular drivers of diastolic HF in pediatric and young adult cardiomyopathy remain poorly defined, where disease is typically driven by primary myocardial disease rather than acquired co-morbidities. The Canadian Cardiomyopathy Collaborative (C3) was assembled to leverage three of Canadas leading pediatric and adult cardiomyopathy biobank registries. Its flagship initiative, Artificial Intelligence to Model Diastolic Heart Failure (AID-HF), aims to integrate deep phenotyping - including comprehensive diastolic function assessment - with genomics, lipidomics and proteomics and apply machine learning to identify biological and clinical signatures that drive cardiac function and outcomes in cardiomyopathy. Harmonized phenotyping and multiomics protocols across registries will create a uniquely integrated national data resource and enable the goals of AID-HF i.e., earlier diagnosis and new therapeutic targets for diastolic HF in cardiomyopathy.

Catheter ablation is an established rhythm-control strategy for atrial fibrillation, but outcomes in persistent atrial fibrillation (PsAF) remain heterogeneous across evolving strategies and energy modalities. An updated synthesis is needed to define current effectiveness and adverse-event profiles in the modern ablation era. We conducted a systematic review and meta-analysis of prospective clinical trials of catheter ablation for PsAF published from 2010 through December 2025. We included randomized and nonrandomized prospective interventional studies reporting effectiveness and adverse events, and pooled outcomes using random-effects models. Prespecified subgroup analyses evaluated ablation strategy (pulmonary vein isolation [PVI] vs PVI with adjunctive lesion sets [PVI+]), ablation modality (radiofrequency [RF], cryoballoon [CRYO], and pulsed field [PF]), and endpoint definition (recurrence-only vs composite measures). Thirty-two studies (9,194 patients) met inclusion criteria; 28 (7,948 patients) contributed to effectiveness analyses. The pooled 12-month arrhythmia-free proportion was 0.65 (95% CI, 0.61-0.68), with substantial heterogeneity. Effectiveness was numerically higher with PVI+ than PVI-only (0.66 [0.60-0.72] vs 0.63 [0.59-0.67]), similar for PF (0.65 [0.57-0.72]) and RF (0.65 [0.61-0.69]), and slightly lower for CRYO (0.64 [0.54-0.74]). Recurrence-only endpoints yielded higher effectiveness than composite endpoints (0.67 [0.63-0.71] vs 0.60 [0.55-0.64]). Safety analyses included 32 studies (9,002 patients). Adverse events were low but heterogeneous (0%-14.56%); pooled vascular access and pericardial complication incidences were each 1%, while thromboembolic events, accessory organ injury, and mortality were rare (pooled 0%). PF ablation showed numerically lower overall complication incidences than RF and CRYO. In contemporary trials, catheter ablation for PsAF shows moderate effectiveness and low overall adverse-event risk. Adjunctive strategies and PF ablation are promising, but no approach is consistently superior. These findings support tailored, patient-specific ablation selection in PsAF.

The Epic Sepsis Model version 2 (ESMv2) is a prediction model embedded into the electronic medical record used to warn clinicians which hospitalized patients are at risk for sepsis. We conducted a retrospective cohort study of 31,951 hospitalizations of 25,760 patients to compare analyses conducted at the commonly used patient-level (where a maximum prediction prior to the onset of sepsis is used to measure performance) vs novel prediction-level (where each prediction is used to measure performance). Sepsis, defined by the Sepsis 3 criteria occurred during 1,049 hospitalizations (3.3%). Patient-level analyses suggested excellent discrimination AUC 0.86; [IQR 0.85, 0.87], whereas prediction-level analyses demonstrated lower performance AUC 0.62; [IQR 0.57, 0.65]. Low estimates of the positive predictive value (14.5% at the patient level vs 4% at the prediction level) imply a high number of false alerts. Common evaluation approaches may overstate the performance of dynamic prediction models and mislead clinical decision-making.

Background Non-vitamin K antagonist oral anticoagulants (NOACs) are the guideline-recommended standard for stroke prevention in atrial fibrillation (AF), yet bleeding risks limit real-world adherence. Percutaneous left atrial appendage closure (LAAC) offers a mechanical alternative without definitive comparative synthesis. Objectives To evaluate percutaneous LAAC versus NOAC therapy by synthesizing all contemporary NOAC-era randomized controlled trials (RCTs). Methods Five databases and registries (PubMed, MEDLINE, Embase, Cochrane CENTRAL, ClinicalTrials.gov) were searched from inception to 8 May 2026 for RCTs comparing percutaneous LAAC against NOACs in adults with non-valvular AF. Risk of bias was assessed using Cochrane RoB 2. Ischemic stroke was pooled using a random-effects DerSimonian-Laird model; primary efficacy composite and non-procedural bleeding were evaluated via pre-specified narrative synthesis. Results Four RCTs (CHAMPION-AF, OPTION, PRAGUE-17, CLOSURE-AF) comprising 5,890 patients were included. LAAC achieved noninferiority for the primary efficacy composite in three trials and demonstrated a statistically significant 45-56% reduction in non-procedural bleeding across the three moderate-risk trials. CLOSURE-AF did not meet noninferiority but retained a directionally consistent bleeding reduction. Pooled ischemic stroke analysis (HR 1.31; 95% CI 0.96-1.80; I^2=0%) showed no statistically significant increase in stroke risk, though a consistent directional trend toward more ischemic events was observed. Conclusions LAAC significantly reduces non-procedural bleeding in moderate-risk AF patients, though this benefit attenuates in very high-risk populations. A consistent, statistically nonsignificant ischemic stroke trend and population-dependent efficacy establish LAAC as a shared decision-making alternative to NOACs rather than a universal replacement, pending 5-year CHAMPION-AF data.

Background: Polygenic risk scores (PRS) for coronary artery disease (CAD) are associated with cardiovascular events, but the relationship between inherited risk and routinely reported coronary computed tomography angiography (CTA) findings has not been studied. Objectives: To evaluate associations between a genome-wide PRS for angiographic coronary disease burden and coronary CTA-derived measures of atherosclerotic severity in a real-world clinical cohort. Methods: We studied Penn Medicine BioBank participants with available genotypes and clinically obtained coronary CTA reports. A previously published PRS for angiographic CAD burden was calculated using pgsc_calc. CAD-RADS scores and coronary artery calcium (CAC) values were extracted from radiology reports using the large language model Llama 3.1 8B. Associations between PRS and CAD-RADS severity were evaluated using Bayesian cumulative ordinal logit regression, while associations with log-transformed CAC burden were assessed using Bayesian linear regression. Results: Among 630 participants, median age was 59 years (IQR 49 - 68), 53% were female, 62% were genetically similar to a European reference population, and 34% to an African reference population. LLM-extracted CAD-RADS and CAC values demonstrated near-perfect agreement with manual abstraction. Higher PRS was associated with greater coronary atherosclerotic burden on CTA. Each 1-standard deviation (SD) increase in PRS was associated with a 20% higher odds of belonging to a more severe CAD-RADS category (cumulative OR 1.20, 95% credible interval 1.06-1.44). Higher PRS was also associated with greater CAC burden ({beta} 0.38, 95% credible interval 0.15 - 0.61). Conclusions: Polygenic risk for angiographic coronary disease burden is reflected in clinically reported coronary CTA severity measures, including CAD-RADS and CAC. These findings demonstrate that inherited susceptibility to CAD manifests as greater anatomic atherosclerotic burden at the time of clinical presentation and support further investigation of genetic risk integration into imaging-based cardiovascular risk assessment.

Importance Dilated cardiomyopathy (DCM) is a major cause of heart failure that disproportionately affects individuals of African genetic ancestry (AFR), among whom familial clustering of disease is also more pronounced relative to those of European ancestry (EUR). However, established monogenic DCM genes, identified primarily in EUR populations, explain a smaller proportion of DCM cases in AFR populations. A recent study identified a common AFR-specific nonsense variant in CD36 that accounts for a substantial burden of DCM in AFR. How the risk and population impact of this variant compare with those of established genetic causes of DCM is unknown. Objective To compare the contribution of a CD36 nonsense variant to DCM risk with that of truncating variants in TTN and pathogenic or likely pathogenic (P/LP) variants in other established DCM genes. Design, Setting, and Participants Multicohort genetic association study including AFR and EUR participants with exome or genome sequence and DCM case status from four datasets: All of Us, Million Veteran Program, Penn Medicine Biobank, and the DCM Precision Medicine Study. Exposure Carrier status for TTN truncating variants, P/LP variants in 11 high confidence DCM genes, and the CD36 nonsense variant (Y325*; 0, 1, or 2 copies). Main Outcomes and Measures Odds of DCM; prevalence of risk-variant carriers among DCM cases; and population attributable fraction (PAF) for DCM. Results Among 82,623 AFR individuals across four studies, the mean age was 53.4 years and 1,625 had DCM. CD36 Y325* risk-allele homozygotes had 4.8-fold (95% CI, 3.1-7.3) increased odds of DCM, and CD36 Y325* heterozygotes had 1.4-fold (95% CI, 1.2-1.7) increased odds. TTN truncating variants also conferred elevated risk of DCM in AFR participants (OR, 8.46; 95% CI, 5.3-12.3). Among AFR DCM cases, 2.5% were CD36 homozygotes, second only to TTN truncating variants (4.3%) and exceeding all other high-confidence DCM genes combined (1.5%). In population-level analyses incorporating both heterozygous and homozygous CD36 Y325* carriers, the population-attributable fraction for CD36 (9.0%) surpassed that of TTN truncating variants (3.6%). Conclusions and Relevance An ancestry-specific CD36 variant contributes more to DCM burden in AFR ancestry than established DCM genes, including TTN truncating variants, typically considered the most common genetic cause of DCM. These findings reshape the known genetic architecture of DCM in individuals of African ancestry and highlight the importance of representation in genomic research.

Background. Peripheral artery disease (PAD) may amplify procedural risk during atrial fibrillation (AF) catheter ablation, but dedicated evidence is lacking. We aimed to evaluate the association between PAD and in-hospital outcomes among adults undergoing AF ablation in the National Inpatient Sample (NIS). Methods. We identified inpatient AF ablation hospitalizations in the 2016 through 2020 National Inpatient Sample using ICD-10-PCS procedure codes and a concurrent AF diagnosis. PAD was identified from ICD-10-CM diagnosis codes used in prior claims-based PAD studies. Stabilized inverse probability of treatment weighting based on the propensity score was used to balance baseline differences. The primary outcome was in-hospital mortality. Fourteen secondary outcomes and 2 composite end points were prespecified. Results. Among 22,166 AF ablation hospitalizations, 899 (4.06%) involved patients with PAD. Compared with patients without PAD, those with PAD were older and had a substantially greater cardiovascular, renal, and smoking/tobacco comorbidity burden. In-hospital mortality did not differ significantly (1.39% vs 1.06%; aOR, 1.32; 95% CI, 0.66 - 2.64; P= 0.44). PAD was associated with higher odds of major bleeding (aOR, 1.62; 95% CI, 1.17 - 2.24; P = 0.004), vascular or access-site complications (aOR, 1.80; 95% CI, 1.04 - 3.12; P = 0.04), acute kidney injury (aOR, 1.31; 95% CI, 1.05 - 1.64; P = 0.02), and composite major adverse hospital events (aOR, 1.29; 95% CI, 1.05 - 1.59; P = 0.02). Total hospital charges were 13% higher (charge ratio, 1.13; 95% CI, 1.04 - 1.22; P = 0.003). Major bleeding, vascular/access-site complications, cardiac arrest, and composite major adverse in-hospital events remained elevated in sensitivity analysis. Conclusion. PAD was independently associated with higher bleeding risk, vascular or access-site complications, acute kidney injury, and composite major adverse hospital event during AF ablation, identifying a clinically relevant subgroup with elevated periprocedural risk.

Acute coronary syndrome (ACS) remains a major contributor to cardiovascular mortality despite advances in emergency cardiovascular intervention and coronary revascularization strategies. This retrospective cohort study evaluated the association between early hemodynamic instability and major adverse cardiovascular events (MACE) among 1,248 ACS patients admitted between January 2023 and December 2025. Patients were categorized into stable and unstable groups based on early emergency department hemodynamic assessment including blood pressure, lactate level, Killip classification, vasopressor requirement, and cardiac output estimation. The primary outcome consisted of 30-day MACE including cardiovascular mortality, recurrent myocardial infarction, cardiogenic shock, ventricular arrhythmia, and urgent revascularization. A total of 372 patients (29.8%) demonstrated early hemodynamic instability and experienced significantly higher rates of cardiogenic shock, ventricular arrhythmia, mechanical ventilation, ICU admission, and 30-day mortality compared with stable patients. Multivariable regression analysis identified serum lactate >4 mmol/L (adjusted OR 3.42; 95% CI 2.10-5.11), systolic blood pressure <90 mmHg (adjusted OR 2.96; 95% CI 1.88-4.47), and left ventricular ejection fraction <35% (adjusted OR 2.71; 95% CI 1.77-4.09) as independent predictors of MACE. Early hemodynamic instability was strongly associated with poor short-term cardiovascular outcomes, suggesting that integrated emergency hemodynamic profiling may improve early risk stratification and facilitate timely cardiovascular intervention.

Background: People with ischemic heart disease (IHD) remain at high risk of recurrent major cardiovascular events despite contemporary therapy. Over two decades, a translational research program has evaluated pressure pain sensitivity (PPS) as a non-invasive marker of central autonomic dysfunction and a mutual risk phenotype in IHD and type 2 diabetes. A PPS-guided non-pharmacological intervention has been shown to substantially reduce five-year all-cause mortality in IHD. Methods: In a randomized controlled trial, 213 adults with stable IHD and elevated PPS, suggesting ANSD, were allocated to PPS-guided intervention (n=106) or control (n=107). The active group received three months of structured education (daily PPS self-measurement, cutaneous sensory nerve stimulation, supportive mental and physical exercises, telemedical feedback) followed by self-directed continuation. Controls received a booklet on general stress-management. The primary endpoint for this prespecified secondary analysis was a composite of eight major cardiovascular events. Results: Over 5 years, at least one major adverse cardiovascular event occurred in 19.8% of the PPS-guided group versus 43.8% of controls (odds ratio 0.32, 95% CI 0.17-0.62, P=0.0003). Incidence rates were directionally in favor of active intervention across all event categories (P=0.004). Conclusions: A brief PPS-guided non-pharmacological intervention, followed by self-directed continuation, was associated with a marked long-term reduction in major adverse cardiovascular events, complementing previously reported large reductions in all-cause mortality in the same cohort. Within the context of a multi-decade PPS research program, these findings support PPS-guided care as a low-resource autonomic intervention ready for pragmatic scale-up testing as an adjunct to cardiometabolic care.

Background: Three-dimensional visualization and quantitative analysis of cerebral arteries on 3DRA are central to endovascular treatment planning, device selection, and cerebrovascular research. Manual segmentation is time-consuming and operator-dependent, yet no open-source deep learning model has been prospectively validated for this task on 3DRA. Methods: A nnUNet v2 model was trained for binary cerebral artery segmentation on 400 consecutive 3DRA acquisitions from three angiographic systems, comparing four configurations across architectures and loss functions. The best-performing configurations were prospectively validated on 40 patients using a dual approach: quantitative metrics (DSC, clDice, HD95, ASD, Precision, Recall), and blinded expert qualitative evaluation by two interventional neuroradiologists assessing 12 arterial segments, a global quality score, and clinical usability across 40 test cases. Results: The ensemble model achieved median DSC 0.917, clDice 0.932, and HD95 1.494 mm. Global quality scores were significantly lower for nnUNet v2 than for expert segmentations (median 4 vs 5, p<0.001), but nnUNet v2 segmentations were rated clinically usable in 88-90% of cases versus 95-98% for expert segmentations, without significant difference on the binary usability criterion. A consistent proximal-to-distal quality gradient was identified, with comparable scores at proximal arteries and the largest differences at distal arterial segments. Conclusion: nnUNet v2 with topology-aware training provides clinically usable cerebral artery segmentations on 3DRA, prospectively validated through both quantitative metrics and structured expert qualitative assessment, and represents a reproducible open-source foundation for endovascular and research applications.

Background: Heterotopic caval valve implantation using the TricValve(R) (OrbusNeich P&F) is a unique interventional approach for treatment of severe Tricuspid Regurgitation in patients who are deemed ineligible for surgery. Given the complexity and novelty of TricValve(R) implantation, there is a pressing need for robust clinical data to evaluate its safety, efficacy, and long-term outcomes. Our study assesses the clinical results of patients followed up for 1 year from our center. Methods: Retrospective, single center registry involving patients who have undergone TricValve(R) Transcatheter Bicaval Valves System (OrbusNeich P&F) implantation for the treatment of severe tricuspid regurgitation. Results: Fourteen patients were included. The mean age was 67.5 {+/-} 8.7 years, with high surgical risk (mean EuroSCORE II 6.1 {+/-} 3.7). Procedural success was achieved in thirteen patients, with no reported in-hospital mortality or stroke among all fourteen patients. At 1-year, significant improvements were observed in New York Heart Association (NYHA) functional class (86% Class III at baseline to 0% Class III at 1 year, P=0.002) and Kansas City Cardiomyopathy Questionnaire (KCCQ-12) scores (mean 32.0 {+/-} 7.4 to 42.4 {+/-} 12.0, P=0.015). TR Regurgitant Volume significantly decreased (65.5 {+/-} 16.9 ml to 38.2 {+/-} 13.6 ml, P=0.005). No deaths or strokes occurred during follow-up. Rehospitalization due to heart failure occurred in 14% (2 out of 14) of patients. Conclusion: In this single-center registry of high-risk patients, TricValve(R) implantation was associated with a favorable safety profile, significant reduction in tricuspid regurgitant volume, and meaningful improvements in functional status and quality of life at 1 year follow-up.

Background & Aims: ACLF is defined differently by APASL (acute hepatic dysfunction) and by organ failure-based frameworks including EASL-CLIF and the recently developed A-TANGO score. Whether these definitions identify competing populations or sequential stages of the same syndrome remains unresolved, with direct implications for the timing of intervention. We tested whether APASL-defined ACLF can be integrated into the A-TANGO framework to identify a clinically actionable patient population. Methods: 4,024 patients hospitalised with acute decompensation of cirrhosis in a multicentre cohort were classified simultaneously by APASL and A-TANGO criteria. Mortality, progression to A-TANGO ACLF among A-TANGO-negative patients, and reversal of ACLF were assessed using Fine-Gray competing-risk models with death as a competing event. EASL-CLIF analyses were performed as sensitivity analyses. Results: A-TANGO-negative/APASL-positive patients comprised 8.7% of the cohort and had higher 90-day mortality than A-TANGO-negative/APASL-negative patients (22.3% vs 14.4%, p=0.001), despite similar 28-day mortality. Once A-TANGO ACLF was established, 28-day mortality was high irrespective of APASL status (45.4% in APASL-positive and 56.0% in APASL-negative patients). Among A-TANGO-negative patients, 53.5% of APASL-positive vs 27.9% of APASL-negative patients progressed to A-TANGO ACLF within 28 days, with APASL positivity independently predicting progression (adjusted sHR: 2.30, 95%CI: 1.90-2.77). Within A-TANGO-negative/APASL-negative patients an A-TANGO OF score [&ge;]8 independently enriched for progression (52% vs 19%). A-TANGO reversal occurred in 17.1% and was independently reduced by APASL positivity (adjusted sHR: 0.756, 95%CI: 0.586-0.975), while APASL reversal was rare (4.0%). EASL-CLIF sensitivity analyses were directionally consistent. Conclusions: APASL-defined ACLF does not compete with A-TANGO; it occupies an upstream position on the same disease trajectory. A-TANGO-negative/APASL-positive patients and A-TANGO-negative/APASL-negative patients with A-TANGO OF [&ge;]8 represent complementary pre-ACLF populations suitable for prevention trials and enrichment strategies.

Background Atrial fibrillation (AF) is a leading cause of stroke, cardiovascular morbidity, and mortality. Atrial myopathy, characterized by progressive metabolic, electrical, and structural changes, creates the arrhythmogenic substrate that drives AF. Defining the key drivers of atrial myopathic processes is essential for targeted therapies that can mitigate AF progression. Here we explore how reduced ERBB4 expression contributes to the development of left atrial myopathy. Methods We analyzed the Cleveland Clinic Biobank to compare left atrial ERBB4 levels in patients grouped by AF diagnosis. To investigate the impact of reduced ERBB4 levels on atrial tissue substrate, we created mouse models of cardiac-specific Erbb4 deficiency using Mlc2a (myosin light chain 2a)-Cre. Comprehensive physiological assessments were performed. Transcriptomic analyses of the left atrium were performed in an Erbb4 haploinsufficient mouse model and compared with human atrial datasets. Molecular validation of key dysregulated pathways was performed. Results We found that left atrial ERBB4 levels are reduced in patients with AF. Adult cardiomyocyte-specific Erbb4 heterozygous (Erbb4fl/+;Mlc2a-Cre) mice exhibited prolonged P-wave duration in the absence of ventricular dysfunction. Left atrial transcriptomic analysis in Erbb4 haploinsufficient mice showed upregulation of pathways related to fibrosis, apoptosis, and coagulation, and downregulation of pathways related to fatty acid metabolism and mitochondrial function, mirroring changes observed in pressure overload mouse models. A cross-species transcriptomic comparison revealed significant overlap between ERBB4-correlated gene expression and functional pathways in adult human atria and mice with Erbb4 haploinsufficiency. Validating the transcriptomic data, protein and functional assays demonstrated increased fibrosis, apoptosis, and oxidative stress in the mutant left atrial tissue. Conclusion Left atrial ERBB4 levels are reduced in AF patients. A mouse model of Erbb4 deficiency and human atrial transcriptomic analyses highlight a role for ERBB4 in supporting normal atrial metabolism while protecting against inflammation, apoptosis, and fibrosis.

Abstract Background Cardiogenic shock (CS) is a heterogeneous syndrome with diverse etiologies, treatment pathways, and outcomes. Prior studies of sex differences in CS have largely focused on acute myocardial infarction-related CS or evaluated CS as a single entity. Whether sex-based differences in outcomes and treatment utilization vary across distinct CS phenotypes remains incompletely defined. Methods We performed a retrospective cohort study using the National Inpatient Sample, a nationally representative all-payer database of United States hospitalizations. Adult hospitalizations with CS were identified using ICD-10-CM code R57.0 and categorized into clinically relevant phenotypes, including acute myocardial infarction (AMI), heart failure (HF), arrhythmia-related shock, myocarditis/Takotsubo, valvular disease, and other etiologies. Survey-weighted analyses accounting for the complex sampling design were used for primary analyses. The primary outcome was in-hospital mortality. Secondary outcomes included use of mechanical circulatory support (MCS) and mechanical ventilation. Propensity score-matched analyses were performed as sensitivity analyses. Results Among 254,691 weighted CS hospitalizations, 158,747 (62.3%) occurred in men and 95,896 (37.7%) in women. In survey-weighted analyses, women had higher in-hospital mortality in AMI-related CS (36.1% versus 31.3%; OR, 1.24; 95% CI, 1.19-1.28), HF-related CS (30.5% versus 25.8%; OR, 1.27; 95% CI, 1.23-1.30), and arrhythmia-related CS (37.3% versus 31.6%; OR, 1.28; 95% CI, 1.20-1.38). Women were less likely to receive ECMO (2.4% versus 2.9%), IABP/Impella (13.1% versus 18.9%), or any MCS (14.6% versus 20.4%), but were more likely to receive mechanical ventilation (44.9% versus 42.9%). In propensity-matched analyses, mortality differences were attenuated but persisted in AMI-related, HF-related, and valvular CS. Conclusions Sex differences in CS outcomes and treatment utilization are strongly phenotype dependent. Women experienced higher mortality in major CS phenotypes while receiving less advanced mechanical circulatory support. These findings support early recognition, rapid phenotype classification, and sex-conscious but non-delayed escalation strategies for women with CS.

Vocal biomarkers, encompassing voice and speech, have largely been developed for individual conditions in isolation, limiting their generalizability across diseases and recording settings. To address this, we introduce VoiceFM, a contrastive model that learns general-purpose clinical voice representations by aligning audio embeddings with rich clinical metadata. Using the Bridge2AI-Voice dataset (984 primarily English-speaking adult participants, 846 used for training and 138 held out as a temporally separated validation cohort, 40,056 recordings totaling 176 hours across 5 academic medical centers), VoiceFM pairs a fine-tuned Whisper large-v2 encoder with a tabular transformer over 44 clinical features via symmetric InfoNCE loss. Linear probes on frozen VoiceFM embeddings achieve mean AUROC 0.952 +/- 0.005 across five evaluation tasks (control vs disease screening plus four disease categories), significantly outperforming Frozen Whisper (0.926 +/- 0.013, p = 0.013), Frozen HuBERT (0.885 +/- 0.017, p = 0.0009), and the contrastively trained VoiceFM-HuBERT (0.938 +/- 0.006, p = 0.012). On the 138-participant held-out cohort, VoiceFM-Whisper achieves AUROCs of 0.99 for Alzheimer's/dementia/MCI and 0.89 for airway stenosis, demonstrating that the learned representations generalize to participants the model has never seen. VoiceFM representations transfer to three external datasets without retraining and improve few-shot classification. Recording task attribution identifies a small set of speech tasks that match or exceed the full battery's performance, suggesting shorter screening protocols are feasible. Trained predominantly on English audio, VoiceFM transfers without fine-tuning to Spanish-language Parkinson's disease (PD) detection (NeuroVoz, 107 participants, AUROC 0.93 +/- 0.02), with the signal dominated by articulatory rather than phonatory features. A fine-tuned classifier achieves participant-level AUROC 0.87 (sustained 0.85, countdown 0.80) on the mPower smartphone study (585 held-out participants). Together, these results show that contrastive alignment between voice and rich clinical metadata can serve as the basis for a clinical voice foundation model, producing a single set of transferable representations that generalize across diseases, languages, recording conditions, and patients enrolled after model freeze.

The electrocardiogram (ECG) encodes the electrical activity of the heart across multiple timescales, yet standard clinical analysis collapses this rich signal into a handful of scalar measurements that discard most of the waveform's structure. Whether the frequency signals lost in this reduction carry heritable biological information relevant to cardiovascular disease risk remains unclear. Here we decompose resting 12-lead ECGs from 47,052 White British UK Biobank participants into 84 frequency-specific energy features using Daubechies-6 wavelet analysis across 12 leads and 7 decomposition levels, and perform independent genome-wide association analyses on each feature. We identify 67 independent loci and refine these to 101 high-confidence causal variants (posterior inclusion probability > 0.80) through Bayesian fine-mapping; associated loci converge on genes governing cardiac conduction and myocardial integrity, including SCN5A, TTN, KCNQ1, and DSP, alongside less-characterized cardiomyopathy candidates. SNP-based heritability estimates range from 0.03 to 0.26, with the strongest signals in mid-frequency bands (D6-D4, ~4-32 Hz) of Lead I and aVR, and strong inter-lead genetic correlations indicate a coordinated genetic architecture underlying the waveform. Integrating these features with FinnGen R12 cardiovascular phenotypes reveals genetic correlations reaching 0.56 with heart failure, driven predominantly by energy in the highest-frequency band (D1, 125-250 Hz), a spectral range routinely filtered from clinical ECGs and previously regarded as acquisition noise. These results reframe the electrocardiogram as a multi-frequency genetic phenotype, expand the set of cardiac loci discoverable from ECG data, and implicate high-frequency cardiac electrical activity as an underexplored dimension of cardiovascular disease risk.

Background and Aims: Clinical interpretation of the precordial leads V1-V6 assumes that Wilson's central terminal (WCT) has a fixed anatomical location. Consequently, a positive signal corresponds to electrical activation spreading from WCT towards the respective electrode, and vice versa. However, the location of WCT has never been systematically investigated. Yet, a better understanding of WCT location could improve the interpretation of the precordial leads. This work aims to characterize the spatial expansion and location of the physical WCT i.e., the electrical potential defined by the WCT, during the P-wave on the body surface. Methods: An intensive analysis of body surface potential maps (BSPMs) during atrial depolarization in an in silico patient cohort and clinical data was conducted. Results: During the P-wave, the location of WCT was not stationary but the spatial extent and location varied across time as well as across individuals. Four distinct spatial patterns of WCT distribution on the body surface were identified in silico, and three of these were found in the clinical cohort. WCT signals agreed with BSPM signals at commonly assumed positions of WCT only for a small fraction of the P-wave. Conclusion: The spatial extension and location of WCT changes during the P-wave and thus should be considered when interpreting the precordial leads.

Objective: Hispanic/Latinx populations in the U.S. experience higher rates of chronic disease linked to physical inactivity, yet digital health interventions remain largely inaccessible to more than 16 million Hispanic/Latinx adults with limited English proficiency. While large language models (LLMs) offer scalable personalization, their use in non-English behavioral coaching is unexplored. This study introduces MHC-Coach-ES, a Spanish-language LLM fine-tuned on the Transtheoretical Model (TTM) of behavior change. Materials and Methods: We fine-tuned Llama 3-70B-Instruct using a two-stage pipeline. First, the model was adapted to Spanish health and motivational language using a 2.21-million-token corpus. Second, it was instruction-tuned on 3,268 translated human written messages to align the model with the Transtheoretical Model (TTM) of Behavioral Change. We compared MHC-Coach-ES with Llama 3-70B-Instruct and translated human-expert messages using a forced-choice preference survey (N = 77) and blinded expert review (N = 2). Results: Spanish-speaking participants significantly preferred MHC-Coach-ES messages over translated human-expert messages (81% preference, P<0.001). Linguistic analysis showed that MHC-Coach-ES produced more temporally anchored messages than the base model (65% vs. 20%), while maintaining readability. In blinded evaluation, clinical experts rated MHC-Coach-ES higher for alignment with Transtheoretical Model stages than human-expert messages (4.83 vs. 4.38 out of 5). The base model also outperformed translated expert messages across preference and expert ratings. Conclusions: Generative AI can operationalize behavioral science frameworks in Spanish, offering a scalable approach to reducing health disparities. The strong performance of both MHC-Coach-ES and the base model highlights the promise of generative and personalized approaches over translation-based localization for theory-driven behavioral interventions.