{
  "abstract": "Background Pulmonary hypertension (PH) is a progressive and underdiagnosed condition associated with substantial morbidity and mortality. Diagnostic delay remains common, partly due to reliance on resource-intensive and operator-dependent investigations such as transthoracic echocardiography. Artificial intelligence (AI)–enabled electrocardiography (ECG) has emerged as a potential low-cost, scalable screening tool for earlier detection of PH. We performed a systematic review and meta-analysis to evaluate the diagnostic accuracy of AI-based ECG models for detecting pulmonary hypertension.Methods This systematic review was conducted in accordance with PRISMA-DTA guidelines and was prospectively registered in PROSPERO (CRD420201241142). MEDLINE (via PubMed), Embase (via Ovid) and Google Scholar were searched using predefined strategies combining terms related to pulmonary hypertension, electrocardiography, and artificial intelligence. Eligible studies included adult populations (≥18 years), applied artificial intelligence, machine learning, or deep learning methods to ECG data for PH detection, and used transthoracic echocardiography and/or right-heart catheterisation as the reference standard. Two reviewers independently screened studies, extracted data, and assessed risk of bias using QUADAS-2. Where sufficient data were available, pooled sensitivity and specificity were calculated using a bivariate random-effects model; otherwise, findings were synthesised narratively.Results Seven studies encompassing over 100,000 ECG recordings were included. Across studies, AI-based ECG models demonstrated good diagnostic performance for the detection of PH. Pooled sensitivity was ≈ 0.81, with a pooled specificity of ≈ 0.81. Reported areas under the receiver operating characteristic curve ranged from 0.84 to 0.88, with a pooled AUC of ≈ 0.89, indicating good discriminative ability. Several studies reported high negative predictive values (>97%), supporting the potential role of AI-ECG as a screening or rule-out tool. Moderate heterogeneity was observed, attributable to differences in study design, AI architecture, ECG format, and reference standards.Conclusions AI-enabled ECG demonstrates good diagnostic accuracy for detecting pulmonary hypertension and may represent a scalable screening tool to facilitate earlier identification and referral for definitive evaluation. Integration of AI-ECG into clinical pathways could reduce diagnostic delay and optimise the use of specialist investigations. Prospective validation and real-world implementation studies are required before routine clinical adoption.",
  "authors": [
    {
      "affiliations": [
        "University of Buckingham, Buckingham, United Kingdom"
      ],
      "name": "Karan Raj Nazareth"
    },
    {
      "affiliations": [
        "University of Buckingham, Buckingham, United Kingdom"
      ],
      "name": "Anisha Sequeira"
    },
    {
      "affiliations": [
        "University of Buckingham, Buckingham, United Kingdom"
      ],
      "name": "Alishba Awais"
    }
  ],
  "title": "556 Diagnostic accuracy of AI-based ECG for pulmonary hypertension: a systematic review",
  "uid": "1931e097-3725-52d0-8234-612133fb6cd0"
}