• HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
Thursday, November 19, 2026
BIOENGINEER.ORG
No Result
View All Result
  • Login
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
No Result
View All Result
Bioengineer.org
No Result
View All Result
Home NEWS Science News Health

AI Electrocardiography Accurately Identifies Ventricular Dysfunction in Left Bundle Branch Block

by
November 19, 2026
in Health
Reading Time: 6 mins read
0
AI Electrocardiography Accurately Identifies Ventricular Dysfunction in Left Bundle Branch Block

AI Electrocardiography Accurately Identifies Ventricular Dysfunction in Left Bundle Branch Block

Share on FacebookShare on TwitterShare on LinkedinShare on RedditShare on Telegram

An artificial intelligence model developed to evaluate routine twelve-lead electrocardiograms can accurately detect left ventricular systolic dysfunction in patients with left bundle branch block while independently predicting long-term risks of mortality and device implantation, according to a retrospective multicenter study conducted across two specialized cardiac hospitals in South Korea. The findings demonstrate that deep-learning analysis of digital electrocardiographic waveforms can overcome longstanding diagnostic hurdles in an electrophysiologically complex patient population, achieving an area under the receiver operating characteristic curve of 0.930. Published in the Journal of Cardiovascular Imaging, the investigation highlights how automated screening might reliably identify patients who warrant prompt echocardiographic assessment and risk-stratify those who face heightened hazards of subsequent adverse cardiac events.

Left bundle branch block represents a distinct conduction disturbance characterized by QRS prolongation and abnormal ventricular activation pathways, which produce mechanical dyssynchrony and impose chronic stress on cardiac mechanics. Because of these detrimental biomechanical effects, individuals presenting with left bundle branch block exhibit a substantially heightened risk of developing left ventricular systolic dysfunction. Prior clinical tracking spanning two to seven years indicates that between 20 percent and 40 percent of these individuals experience meaningful declines in left ventricular function over time. However, conventional electrocardiography cannot determine whether systolic impairment is present or gauge its severity, because the characteristic broad QRS complex, intrinsic conduction delay, and secondary repolarization abnormalities obscure standard electrocardiographic signs of myocardial dysfunction.

Although transthoracic echocardiography remains the established clinical reference standard for evaluating left ventricular ejection fraction, access to routine imaging is frequently constrained by healthcare costs, institutional capacity, and the necessity for specialized operator expertise. Even when patients present with an initially preserved ejection fraction, subsequent deterioration can emerge unpredictably, creating an urgent clinical demand for scalable monitoring tools that do not rely on immediate echocardiography. While artificial intelligence models analyzing twelve-lead electrocardiograms have shown robust diagnostic performance across general clinical populations, their diagnostic reliability specifically within left bundle branch block had remained largely unexamined due to historical concerns that severe baseline conduction distortions might degrade algorithmic accuracy.

To rigorously investigate this capability, researchers evaluated 5,689 expert-validated electrocardiograms obtained from 2,813 adult patients who presented with confirmed left bundle branch block and underwent paired echocardiography within fourteen days. Data collection spanned January 2016 through December 2024 at Bucheon Sejong Hospital and January 2017 through December 2024 at Incheon Sejong Hospital, both government-designated cardiac specialty institutions. To ensure high diagnostic certainty, initial electrocardiograms displaying potential left bundle branch block were flagged using the automated interpretation software of the MUSE Cardiology Information System and subsequently reviewed manually by an electrocardiography specialist. Patients aged eighteen years or older were included, establishing a curated cohort with a mean age of 70.7 years and a male representation of 43.7 percent.

Standard twelve-lead electrocardiograms were recorded for ten seconds at a sampling rate of 500 hertz using Philips PageWriter devices before storage in the institutional database. Automated feature extraction captured intrinsic electrical variables, including heart rate, PR interval, QRS duration, QT interval, corrected QT interval, and electrical axes. Concurrently, electronic health records provided comprehensive clinical profiles, including body mass index, blood pressure measurements, and key comorbidities such as hypertension, diabetes mellitus, atrial fibrillation, ischemic heart disease, chronic kidney disease, and documented heart failure. Transthoracic echocardiograms yielded biplane Simpson ejection fractions alongside structural and hemodynamic metrics, including global longitudinal strain, chamber dimensions, filling pressure ratios, and valvular regurgitation, with all-available case analyses applied without statistical imputation.

The study evaluated the AiTiALVSD model, a deep-learning algorithm combining convolutional neural network architectures with transformer layers to directly process raw digital ten-second twelve-lead waveforms without manual feature extraction. Originally developed on a multi-institutional training set of 449,586 electrocardiograms from 167,678 patients across sixteen medical centers, the software generates an output score from 0 to 100 denoting the probability of left ventricular systolic dysfunction. The algorithm received medical device authorization from the Korean Ministry of Food and Drug Safety in 2024. In the present study, all validation cases from Bucheon Sejong Hospital were strictly excluded from prior model development cohorts, and Incheon Sejong Hospital contributed no developmental data, ensuring complete patient-level and temporal independence between training and testing populations.

Across the validation cohort, 1,157 patients, representing 41.1 percent of the population, exhibited left ventricular systolic dysfunction, defined as an ejection fraction of 40 percent or lower measured according to American Society of Echocardiography guidelines. Patients with systolic dysfunction demonstrated higher prevalences of diabetes, ischemic heart disease, chronic kidney disease, and prior heart failure, alongside lower baseline systolic blood pressure and more pronounced conduction prolongation, with an overall mean QRS duration of 153.4 milliseconds. When analyzing digital waveforms, the model produced average probability scores of 69.4 in patients with confirmed dysfunction compared to 18.2 in those with preserved ejection fraction, reflecting clear electrophysiological discrimination between clinical subgroups.

Diagnostic assessment revealed that the model achieved an area under the receiver operating characteristic curve of 0.930 and an area under the precision-recall curve of 0.913 for identifying left ventricular systolic dysfunction. Applying a predefined decision threshold derived to achieve approximately 90 percent sensitivity in the model development dataset, the algorithm yielded a sensitivity of 0.979, a specificity of 0.473, a positive predictive value of 0.594, and a negative predictive value of 0.967 in the validation population. The elevated negative predictive value indicates that low scores can effectively rule out systolic impairment in this population, supporting a triage approach that safely defers immediate echocardiography in low-risk individuals while directing urgent imaging resources toward patients flagged with high-risk probability scores.

Model performance remained stable across alternative clinical thresholds and diagnostic classifications. When evaluated against a lower ejection fraction cutoff of 35 percent, which informs decisions regarding defibrillator implantation, the algorithm achieved an area under the curve of 0.927, paired with 99.0 percent sensitivity, 42.7 percent specificity, and a negative predictive value of 98.7 percent. When assessed against an ejection fraction threshold of 50 percent, the area under the curve reached 0.907, with 93.8 percent sensitivity and 58.3 percent specificity. Sensitivity analyses across probability cutoffs ranging from 20 to 90 demonstrated predictable trade-offs between sensitivity and specificity, while subgroup analyses revealed consistent discriminative performance across patient age, sex categories, diabetes status, and degrees of QRS prolongation.

Beyond binary classification, probability scores exhibited a strong inverse correlation with continuous echocardiographic ejection fraction measurements, documented by a Pearson correlation coefficient of minus 0.81. Higher model scores clustered among individuals with severe mechanical impairment, whereas lower scores aligned with preserved ventricular function. Saliency mapping utilizing gradient-based computations provided preliminary insights into the electrical features driving model predictions. High-risk electrocardiograms displayed pronounced focal activation within lead V1, concentrated specifically around the PR-interval segment, suggesting that subtle atrioventricular conduction delays and early septal activation patterns provide meaningful predictive indicators, whereas low-risk maps demonstrated more diffuse and less intense signal distributions across leads.

During a mean longitudinal follow-up duration of 4.1 years, risk stratification based on the model’s predefined threshold demonstrated substantial prognostic value. Patients categorized into the high-risk tier experienced significantly greater hazards for all-cause mortality, yielding an adjusted hazard ratio of 1.87 with a 95 percent confidence interval of 1.53 to 2.28 in multivariable models accounting for age, sex, and baseline comorbidities. High-risk classification was also associated with marked elevations in implantable cardioverter defibrillator or cardiac resynchronization therapy implantation, showing an adjusted hazard ratio of 15.2. Among patients with preserved baseline systolic function, high-risk status predicted an adjusted hazard ratio of 12.05 for incident reductions in ejection fraction to 40 percent or less, while cardiovascular hospitalizations demonstrated an adjusted hazard ratio of 1.11 that did not attain statistical significance.

The authors emphasized several methodological considerations and study limitations. Because the cohort was assembled retrospectively from patients who underwent both confirmed electrocardiography and echocardiography within a two-week interval, the study is subject to selection bias, resulting in a higher observed prevalence of systolic dysfunction and heart failure than would occur in an unselected screening population. Consequently, performance metrics reflect an enriched hospital-based cohort rather than community-based screening. Additionally, the investigation was conducted within two medical centers in the Republic of Korea, indicating the need for prospective validation across ethnically diverse cohorts. The authors noted that the model functions as a complex deep-learning system and that the observational design could not capture dynamic therapy optimizations over time or provide formal economic evaluations.

Subject of Research: Artificial intelligence evaluation of left ventricular systolic dysfunction and long-term prognosis in left bundle branch block electrocardiograms

Article Title: Detection and prognostic stratification of left ventricular systolic dysfunction in left bundle branch block using an artificial intelligence–enabled electrocardiography

Article References: Lee, S. Y., Yoo, A.-H., Kang, S., Jang, J.-H., Jo, Y.-Y., Son, J. M., Lee, M. S., Han, G. I., Kwon, J.-M., Lee, H. S., & Kim, K.-H. (2026). Detection and prognostic stratification of left ventricular systolic dysfunction in left bundle branch block using an artificial intelligence–enabled electrocardiography. Journal of Cardiovascular Imaging, 34(1), Article 4. https://doi.org/10.1186/s44348-026-00066-9

Image Credits: AI Generated

DOI: 10.1186/s44348-026-00066-9

Keywords: artificial intelligence, deep learning, electrocardiography, left bundle branch block, left ventricular systolic dysfunction, ejection fraction, cardiac resynchronization therapy, implantable cardioverter defibrillator, diagnostic accuracy, prognostic stratification, all-cause mortality, saliency mapping, multicenter validation, digital biomarker, echocardiography triage, cardiovascular risk

News Source: Ophelia Keating. (November 19, 2026). AI Electrocardiography Accurately Identifies Ventricular Dysfunction in Left Bundle Branch Block. Scienmag.

Tags: all-cause mortalityArtificial Intelligencecardiac resynchronization therapycardiovascular riskdeep learningdiagnostic accuracydigital biomarkerechocardiography triageejection fractionelectrocardiographyimplantable cardioverter-defibrillatorleft bundle branch blockleft ventricular systolic dysfunctionmulticenter validationprognostic stratificationsaliency mapping
Share12Tweet7Share2ShareShareShare1

Related Posts

UK Launches National Consortium to Transform Population Mental Health Prevention and Policy

UK Launches National Consortium to Transform Population Mental Health Prevention and Policy

November 19, 2026
Climate Hazards and Recurrent Floods Linked to Heightened Depression Among Thai Farmers

Climate Hazards and Recurrent Floods Linked to Heightened Depression Among Thai Farmers

November 19, 2026

Histone Demethylase KDM6B Regulates Adipose Tissue Expansion by Restraining Perivascular Macrophages

November 19, 2026

Preoperative Knee Alignment and Contracture Predict Polyethylene Insert Thickness in Arthroplasty

November 19, 2026

POPULAR NEWS

  • Review Outlines Extraction and Health Benefits of Blackcurrant Polysaccharides

    Review Outlines Extraction and Health Benefits of Blackcurrant Polysaccharides

    29 shares
    Share 12 Tweet 7
  • Canadian Focus Groups Reveal Core Lessons for Wildlife One Health Integration

    29 shares
    Share 12 Tweet 7
  • Bevacizumab Biosimilar Shows Similar Blood Clot Risk to Reference Drug in Cancer Study

    29 shares
    Share 12 Tweet 7
  • Biotelemetry Array Identifies Wild Murray Cod Nesting Sites in Turbid Waters

    29 shares
    Share 12 Tweet 7

About

We bring you the latest biotechnology news from best research centers and universities around the world. Check our website.

Follow us

Recent News

UK Launches National Consortium to Transform Population Mental Health Prevention and Policy

Climate Hazards and Recurrent Floods Linked to Heightened Depression Among Thai Farmers

Histone Demethylase KDM6B Regulates Adipose Tissue Expansion by Restraining Perivascular Macrophages

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 90 other subscribers
  • Contact Us

Bioengineer.org © Copyright 2023 All Rights Reserved.

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • Homepages
    • Home Page 1
    • Home Page 2
  • News
  • National
  • Business
  • Health
  • Lifestyle
  • Science

Bioengineer.org © Copyright 2023 All Rights Reserved.