Philipp Sebastian Wild

Aims: Artificial intelligence, particularly deep learning (DL), has shown significant potential for enhancing the assessment of echocardiograms in clinical practice. This systematic review and meta-analysis aimed to evaluate the performance of DL models and to identify factors affecting their performance. Methods and results: A systematic review of the PubMed, CENTRAL, and Web of Science databases was conducted, yielding 16 248 potential publications. Of these, 160 studies met the inclusion criteria. Only studies applying fully automated DL networks to echocardiographic loops were considered. Performance estimates were pooled using a random-effects meta-analysis. The studies were categorized into five distinct groups: prediction of disease or mortality, replication of established measures, segmentation of cardiac chambers, classification of chamber views, and assessment of image quality. High heterogeneity that could not be resolved was observed in studies focusing on cardiac chamber segmentation and classification. Model performance was influenced by a variety of different factors, including medical target and input characteristics such as video resolution, number of frames used, and total participants included during training. Conclusion: DL models achieve high performance across various applications in echocardiography. In selected tasks, they may reduce reliance on expert input and support greater automation of echocardiographic workflows. However, further validation, assessment of clinical utility, and evaluation in real-world settings are required before widespread clinical adoption.

Background: A patient undergoes multiple examinations in each hospital stay, where each provides different facets of the health status. These assessments include temporal data with varying sampling rates, discrete single-point measurements, therapeutic interventions such as medication administration, and images. While physicians are able to process and integrate diverse modalities intuitively, neural networks need specific modeling for each modality complicating the training procedure. Methods: We demonstrate that this complexity can be significantly reduced by visualizing all information as images along with unstructured text and subsequently training a conventional vision-text transformer. Our approach, Vi sion T ransformer for i rregular sampled M ulti-modal M easurements ( ViTiMM ), simplifies data preprocessing and modeling by unifying clinical measurements, medications, X-ray images, and electrocardiography scans into a single visual representation. Results: ViTiMM outperforms current state-of-the-art methods in predicting in-hospital mortality, phenotyping, and decompensation on two datasets, the MIMIC-IV and COVID Data for Shared Learning (CDSL) dataset. Conclusions: We hope our work inspires advancements in multi-modal medical AI by reducing the training complexity to (visual) prompt engineering, thus lowering entry barriers and enabling no-code solutions for training. The source code is publicly available at https://github.com/Cardio-AI/ViTiMM .

Key summary points: Aim: To investigate whether older individuals (≥70 years) with heart failure receive guideline-recommended heartfailure medications less frequently than younger individuals. Findings: Older individuals had a higher prevalence of heart failure and greater polypharmacy but were treated lessoften with guideline-based therapies, including aldosterone antagonists and beta-blockers, even after adjustment forrenal function and comorbidities. In contrast, calcium channel blockers were prescribed more frequently in geriatricpatients. Message: Older individuals with heart failure appear to be undertreated with evidence-based therapies, highlightingage-related gaps in guideline-concordant care. Background: In Germany, heart failure (HF) is the most common diagnosis in hospitalized patients. Its prevalence rises significantly with age. In addition, age-related conditions, such as hypertension, coronary artery disease, obesity, and diabetes mellitus, contribute to an elevated risk of heart failure. It is assumed that older patients are more likely to receive treatment that does not conform to the guidelines because of frequent comorbidities, such as chronic renal disease and pre-existing polypharmacy. Methods: Data from a prospective cohort study on chronic HF (MyoVasc study, N = 3289; NCT04064450), were analyzed. Participants underwent a 5-h highly standardized examination, including assessment of medication with subsequent categorization according to the anatomical therapeutic chemical (ATC) coding system. Study participants were categorized as older adults aged ≥ 70 years and subsequently compared to younger adults regarding frequency of medication intake. Poisson regression with robust standard errors with adjustment for age, sex, cardiovascular risk factors, comorbidities, and kidney function was used to analyze medication intake by age status in subjects with chronic HF stages C/D. Results: The analyzed sample comprised 1281 older adults (mean age: 75.4 ± 3.6 years, 33.1% women, 66.9% men) and 2,008 younger adults (mean age: 57.7 ± 8.5 years, 35,2% women, 64.8% men). A total of 68.5% ( n = 878) of the older individuals presented with HF stage C/D, compared to 43% ( n = 863) in younger adults. On average, older individuals took 6.53 (standard deviation (SD) ± 3.14) drugs as compared to younger participants taking 4.65 (SD ± 3.45) drugs regardless of the severity of HF. Poisson regression analysis demonstrated that older individuals received aldosterone antagonists less frequently (Prevalence Ratio (PR): 0.64, 95%CI 0.54; 0.77, p < 0.0001). This effect persisted after additional adjustment for kidney function (PR: 0.53, 95%CI 0.44; 0.63, p < 0.0001). Beta-blockers were used significantly less frequently in older adults compared to younger adults (PR: 0.93, 95%CI 0.88; 0.99, p = 0.016). In contrast, calcium channel blockers were prescribed significantly more often in the older group (PR: 1.22, 95%CI 1.01; 1.48, p = 0.035). Additionally, older adults with impaired renal function received loop diuretics less frequently (PR: 0.88, 95%CI 0.78; 0.99, p = 0.033). Conclusion: The present study identified age-related disparities in the use of guideline-based therapy for heart failure. These differences persisted even after adjusting for renal dysfunction, a common and relevant comorbidity in the aged population. These findings suggest a potential underutilization of evidence-based heart failure therapies in older individuals with heart failure. Trial registry: NCT04064450

Aims: The aim of the project was to better define chronotropic incompetence (CI) and evaluate its clinical determinants and relation to clinical outcomes in heart failure (HF). Methods: Data from the MyoVasc study (NCT04064450) were analysed. Subjects underwent deep clinical phenotyping, including cardiopulmonary exercise testing (CPET). Individuals with a respiratory exchange ratio < 1.0 were excluded. Continuous predicted heart rate reserve (pHRR) operationalized CI. A reference range for CI was derived from a healthy subsample not taking β-blockers. Determinants of CI were assessed with a bidirectional stepwise approach and clinical outcomes by Cox regressions. Results: The analysis sample included 880 subjects [median age 63 years (interquartile range 55; 72); 30.0% women]. Symptomatic HF was present in 42.8% (377) of subjects and 64.7% (n = 569) were considered chronotropic incompetent based on the traditional 80% threshold. According to the new reference range, CI was defined as a pHRR < 56.4%. Peak oxygen uptake (VO2), β-blockers, age, sex, smoking, CPET protocol, HF severity, C-peptide, and a fatty liver were among the CI determinants. No phenotype-specific pattern of CI determinants was observed. In Cox regression analysis, CI independently predicted all-cause death [hazard ratio (HR) per standard deviation (SD) decrease in pHRR (HRSD) 1.77 (1.42; 2.21); P < .0001], cardiac death [HRSD 1.74 (1.08; 2.80); P = .023], and worsening of HF (HRSD 1.39 [1.01; 1.91], P = .046). Conclusion: Chronotropic incompetence was a strong predictor of clinical outcomes. Its associations with hyperinsulinemia and abnormal liver function underlined the connection between metabolism and CI. This highlights the relevance of impaired response to exercise in the clinical course of HF with implication for risk stratification.

Background: While proteomic alterations in severe COVID-19 have been described, the molecular differences that distinguish asymptomatic from symptomatic infection and their relationship to long-term outcomes remain incompletely understood. Understanding these mechanisms may help explain inter-individual differences in disease severity and post-COVID sequelae.Methods: Plasma samples were analyzed from individuals with (N=494) and without (N=190) SARS-CoV-2 infection using mass spectrometry-based proteomics in a longitudinal population-based cohort pre- and post-infection. Ensemble feature selection and machine learning approaches were used to identify proteins associated with symptomatic versus asymptomatic infection.Findings: The model identified 49 proteins associated with symptomatic infection (10-fold cross-validated AUC=0.85). Proteins associated with symptomatic infection were involved in complement, coagulation, and inflammatory pathways, whereas asymptomatic infection was associated with lipoprotein metabolism and immune-metabolic pathways. A protein score derived from this signature was associated with acute symptom burden, disease severity, and post-COVID sequelae. Plastin-2 (LCP1) and complement factor B (CFB) were associated with symptomatic infection and multiple acute symptoms, whereas Apolipoprotein D (APOD) was associated with asymptomatic infection and fewer symptoms. Higher pre-infection HDL concentrations were inversely associated with the symptomatic protein score.Interpretation: Preservation of lipoprotein-related proteins were associated with asymptomatic SARS-CoV-2 infection, whereas complement and coagulation activation with symptomatic infection. These findings identify an immune-metabolic axis linking lipid metabolism, complement activation, and COVID-19 severity and may contribute to biomarker-based risk stratification and understanding of post-COVID outcomes.

Background: While proteomic alterations in severe COVID-19 have been described, the molecular differences that distinguish asymptomatic from symptomatic infection and their relationship to long-term outcomes remain incompletely understood. Understanding these mechanisms may help explain inter-individual differences in disease severity and post-COVID sequelae.Methods: Plasma samples were analyzed from individuals with (N=494) and without (N=190) SARS-CoV-2 infection using mass spectrometry-based proteomics in a longitudinal population-based cohort pre- and post-infection. Ensemble feature selection and machine learning approaches were used to identify proteins associated with symptomatic versus asymptomatic infection.Findings: The model identified 49 proteins associated with symptomatic infection (10-fold cross-validated AUC=0.85). Proteins associated with symptomatic infection were involved in complement, coagulation, and inflammatory pathways, whereas asymptomatic infection was associated with lipoprotein metabolism and immune-metabolic pathways. A protein score derived from this signature was associated with acute symptom burden, disease severity, and post-COVID sequelae. Plastin-2 (LCP1) and complement factor B (CFB) were associated with symptomatic infection and multiple acute symptoms, whereas Apolipoprotein D (APOD) was associated with asymptomatic infection and fewer symptoms. Higher pre-infection HDL concentrations were inversely associated with the symptomatic protein score.Interpretation: Preservation of lipoprotein-related proteins were associated with asymptomatic SARS-CoV-2 infection, whereas complement and coagulation activation with symptomatic infection. These findings identify an immune-metabolic axis linking lipid metabolism, complement activation, and COVID-19 severity and may contribute to biomarker-based risk stratification and understanding of post-COVID outcomes.