European heart journal. Digital health最新文献

[This corrects the article DOI: 10.1093/ehjdh/ztaf143.150.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.151.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.152.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.153.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.154.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.155.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.156.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.157.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.158.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.159.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.160.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.161.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.162.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.163.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.164.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.165.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.166.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.167.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.168.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.169.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.170.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.171.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.172.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.173.][This corrects the article DOI: 10.1093/ehjdh/ztaf143.174.].

Aims: Artificial intelligence (AI)-based electrocardiogram (ECG) analysis tools have shown promise in detecting various cardiac conditions. However, their performance in specific patient populations, such as those with hypertrophic cardiomyopathy (HCM), remains incompletely characterized. To evaluate the performance of three AI-based ECG analysis tools in patients with confirmed HCM: (1) a tool calculating HCM probability, (2) a tool calculating structural heart disease (SHD) probability, and (3) a tool providing ECG-based diagnoses across multiple categories.

Methods and results: We analysed digitized 12-lead ECGs from patients with confirmed HCM (n = 681) using three AI tools. We assessed the distribution of AI-calculated probabilities and their associations with clinical parameters and evaluated agreement between AI-based and manually assigned ECG diagnoses using Cohen's kappa. Despite all patients having confirmed HCM, the AI-calculated HCM probabilities showed a relatively uniform distribution [median 38.8% (IQR: 12.8-63.4%)], with only 41.2% and 12.5% of patients receiving a probability score >50% and >75%. HCM probabilities were significantly higher in patients with abnormal vs. normal ECGs (P < 0.001) and correlated with markers of disease severity. SHD probabilities were generally higher [median 51.4% (IQR: 28.7-74.5%)], with 51.2% and 25% of patients receiving scores >50% and >75%.

Conclusion: AI-based ECG analysis tools demonstrated modest performance in our HCM cohort. These findings highlight the challenges of applying AI tools developed in general populations to specific disease cohorts and underscore the need for disease-specific validation before clinical implementation.

Aims: We applied unsupervised machine learning clustering to a large cohort of hypertensive patients undergoing echocardiography with strain imaging to identify phenotypes with distinct clinical profiles, comorbidities, remodelling trajectories, and outcomes.

Methods and results: We analysed 1607 patients from the STRATS-HHD registry who underwent echocardiography at baseline and after 6-18 months of therapy. Twenty clinical, laboratory, and echocardiographic variables-including left atrial and left ventricular strain-underwent principal component analysis and K-means clustering (K = 4). Clusters were derived in the SNUBH cohort (n = 1204) and validated in the CAUH cohort (n = 403), two institutional subsets of the registry. Remodelling trajectories were assessed using baseline-adjusted models, and associations with outcomes were evaluated using multivariable Cox regression. Four clusters emerged: (i) atrial fibrillation-predominant, with advanced remodelling and the highest event risk; (ii) elderly, with metabolic-renal comorbidities but preserved function; (iii) middle-aged, with prevalent coronary disease and relatively preserved function; and (iv) younger, with severe hypertension, marked strain impairment, and the greatest remodelling regression with therapy. Prognosis varied: cluster 1 had the highest risk of cardiovascular death, heart failure hospitalization, stroke, and major adverse cardiovascular events (MACE); cluster 2 exhibited increased cardiovascular death and intermediate heart failure hospitalization risk; cluster 3 showed elevated coronary risk; and cluster 4 the most favourable outcomes. Associations between medication and remodelling varied, with renin-angiotensin blockade linked to LV mass regression in cluster 4.

Conclusion: Machine learning -based clustering incorporating strain identified four distinct HHD phenotypes with divergent remodelling, therapeutic responses, and outcomes. Data-driven phenotyping may improve risk stratification and enable tailored management in hypertension.

Aims: Valvular heart disease (VHD) can affect more than one in two adults over 65 yet remains underdiagnosed due to the limitations of traditional auscultation. Earlier detection is critical for improving outcomes, but many cases go unrecognized. This study evaluates whether an artificial intelligence (AI)-enabled digital stethoscope can augment primary care providers' (PCPs) ability to detect clinically significant VHD compared with analogue auscultation alone.

Methods and results: In this prospective study, 357 patients aged ≥50 years at risk for heart disease underwent both analogue cardiac auscultation by PCPs (standard of care, SOC) and digital cardiac auscultation by study coordinators followed by AI analysis using an electronic stethoscope (AI-augmented). Echocardiography and audible murmur annotation served as the reference standard. Sensitivity and specificity of AI vs. SOC were compared using Fisher's exact test. The AI-augmented system demonstrated significantly higher sensitivity (92.3% vs. 46.2%, P = 0.01) but lower specificity (86.9% vs. 95.6%, P < 0.001) compared with SOC. Artificial intelligence detected 12 cases of previously undiagnosed mod+ VHD, while routine auscultation identified 6.

Conclusion: Artificial intelligence-enabled digital stethoscopes significantly improve point-of-care VHD detection, offering a promising tool for earlier diagnosis and intervention in primary care settings.

Aims: Acute kidney injury (AKI) is a frequent and severe complication in critically ill patients with cardiovascular instability. Current risk scores rely on delayed renal biomarkers such as serum creatinine (sCr) and blood urea nitrogen (BUN). We aimed to develop and validate machine learning (ML) models predicting AKI and major adverse kidney events (MAKE) exclusively from systemic physiological and haemodynamic data.

Methods and results: Two ML models were trained on the MIMIC-IV database: one including (sCr+/BUN+) and one excluding (sCr-/BUN-) renal parameters. External validation was performed in the eICU database and in a cohort of burn ICU patients from AP-HP. Model performance was assessed for early AKI and MAKE prediction up to 100 h before diagnosis. Systemic haemodynamic and physiological variables were the strongest predictors of AKI. In MIMIC-IV, the sCr-/BUN- model achieved auROC 0.78 at 72 h, approaching the sCr+/BUN+ model. In eICU, it outperformed the biomarker-based model at later time points (auROC 0.73). In the burn ICU cohort-representing a high-stress systemic environment-it maintained robust accuracy (auROC 0.75 at 24 h, 0.77 at 72 h). For MAKE prediction, the sCr-/BUN- model achieved auROC 0.87 (burn cohort), 0.67 (eICU), and 0.77 (MIMIC-IV). Median lead time for AKI prediction exceeded 70 h.

Conclusion: AI models based solely on non-renal parameters can accurately predict AKI and MAKE, even under extreme systemic stress such as severe burns. Haemodynamic signatures carry sufficient information to anticipate kidney dysfunction well in advance, opening the way to real-time, proactive cardio-renal risk stratification in ICU patients with acute heart failure, cardiogenic shock, and after cardiac surgery.

Aims: To develop a deep-learning (DL) framework that enables fully automated longitudinal and circumferential co-registration of intravascular ultrasound (IVUS) and optical coherence tomography (OCT) images.

Methods and results: Data from 230 patients (714 vessels) with acute myocardial infarction that underwent near-infrared spectroscopy IVUS and OCT imaging in their non-infarct related vessels were analysed. Experts annotated the lumen borders (61 655 IVUS and 62 334 OCT frames), the side branches and the calcific tissue (10 000 IVUS and 10 000 OCT frames each). This information was used to train DL models that extracted these features that were then used by a dynamic time warping algorithm to co-registered longitudinally the IVUS and OCT images. The circumferential registration of IVUS and OCT was performed through a rotation cost matrix and dynamic programming. On a test set of 22 patients (77 vessels), the DL method showed high concordance with the expert analysts for the longitudinal and circumferential co-registration of the two datasets (concordance correlation coefficient >0.99 and >0.90, respectively). The Williams Index was 0.96 for longitudinal and 0.97 for circumferential alignment, indicating a comparable performance of the proposed framework to the analysts. The time needed for the DL pipeline to process imaging data from a vessel was <90 s.

Conclusion: A fully automated, DL-based framework for IVUS-OCT co-registration demonstrated both speed and accuracy, with performance comparable to that of expert analysts. These features enable its application in research using large-scale data incorporating multimodality imaging.

Aims: Existing ST-segment elevation myocardial infarction (STEMI) alert pathways that rely on traditional STEMI criteria perform suboptimally. We aimed to evaluate the diagnostic performance of an artificial intelligence (AI) model to detect acute occlusion myocardial infarction (OMI) from the routine 12-lead electrocardiogram (ECG) and, specifically, its potential to reduce false-positive activations.

Methods and results: Consecutive adults managed via the STEMI pathway were included from a tertiary academic medical centre between January 2022 and December 2023. Cases without an available ECG for review, death prior to catheterization, or alternative reasons for activation (i.e. electrical instability or urgent interventions) were excluded. Pre-coronary angiography tracings were interpreted via the AI tool. Test characteristics were compared against traditional STEMI criteria. The primary outcome was the number of avoidable false-positive activations. During the 2-year study period, there were 454 activations, 150 were excluded, and 304 cases with unique ECGs were included in the study cohort. There were 118 (38.8%) false-positive activations, of which 86 (72.9%) were correctly predicted by the AI model. Its test characteristics for identifying true positives were superior compared with traditional STEMI criteria for a sensitivity of 89.2% [95% confidence interval (CI): 84.0-92.9] vs. 68.3% (95% CI: 61.3-74.5), specificity 72.9% (95% CI: 64.2-80.1) vs. 51.7% (95% CI: 42.8-60.5), and accuracy 82.9% (95% CI: 78.3-86.7) vs. 61.8 (95% CI: 56.3-67.1).

Conclusion: The AI model is superior to traditional STEMI criteria for detecting OMI and has the potential to reduce false-positive catheterization lab activations. It can be a useful decision-aid for catheterization lab activation.

Aims: Accurate diagnoses contribute to the improvement of clinical workflows and the enhancement of patient care. Commercially available automated electrocardiogram (ECG) interpretation systems require manual review by physicians despite their widespread use. This study investigates the frequency and characteristics of the modifications from automated ECG reports in routine clinical practice.

Methods and results: We retrospectively analysed 159 630 ECGs from 2011 to 2023 and compared automated preliminary ECG reports generated by the GE Marquette™ 12SL ECG analysis programme with finalized reports by physicians. A modification was defined as any textual difference between the initial and final reports. Our analysis revealed that 31.3% of all ECG reports underwent some forms of modification by physicians. We analysed the frequency of 69 pre-defined ECG-related terms before and after physician review, categorizing modifications as unchanged, deleted, or newly added. Modifications were more frequent for ECGs performed during off-hours, in patients with higher ventricular rates and longer QRS durations. At the term-level, diagnoses such as 'prolonged QT interval' (newly added from 5.6% of original reports) and 'electronic ventricular pacemaker' (newly added from 3.6% of original reports) were frequently added by physicians, while diagnoses like 'inferior infarct' and 'anterior infarct' were frequently deleted from automated ECG reports (32.0% and 44.6% automated reports with these terms required removals).

Conclusion: This large-scale real-world study demonstrated the high frequency of physicians' modification in automated ECG interpretation. The identified patterns of modifications highlight the limitations of current rule-based systems in handling complex cases and nuanced ECG findings.

Aims: Unenhanced chest CT is frequently used to assess lung malignancy and parenchymal disease. Harnessing CT data to quantify cardiac and vascular structures has the potential to improve the diagnosis of heart failure and pulmonary hypertension (PH). This study aims to develop a deep learning model to segment and analyse cardiothoracic structures from unenhanced CT images to diagnose PH, pre-capillary PH and PH associated with left heart disease (LHD).

Methods and results: A twelve-structure cardiothoracic segmentation model was developed using an institutional cohort (n = 55, 35/9/11 training/validation/testing). Model performance was evaluated using Dice similarity coefficients (DSC). Volumetric measurements were compared to manual values using intra-class correlation (ICC) and visually assessed by four observers using an external cohort (n = 50, from 26 hospitals). Univariable and multivariable regression analyses were performed using a cohort of 368 patients (254/114 training/testing). Receiver-operating characteristic curves were plotted and the area under the curves (AUC) with confidence intervals (CI) were calculated. The model yielded a DSC segmentation performance of ≥0.87 for 9/12 segmented structures and ICC > 0.95 for 10/12 structures. Most of the segmented structures scored as excellent in the external cohort visual assessment. Diagnostic accuracy for predicting PH was high [AUC = 0.88 (CI: 0.80-0.96), sensitivity = 70%, specificity = 100%], including pre-capillary PH [AUC = 0.84 (CI: 0.74-0.94), sensitivity = 72%, specificity = 94%] and PH-LHD [AUC = 0.86 (CI: 0.79-0.93), sensitivity = 94%, specificity = 63%].

Conclusion: A fully automated model for multi-structure cardiothoracic segmentation on unenhanced CT is achievable. The model can predict PH and identify patients with pre-capillary PH and PH-LHD with promising performance.