Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
August 30, 2026Discover Artificial IntelligenceOpen Access

Machine learning models for ECG-based CVD prediction commonly exceed ~90% accuracy.

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Recent machine learning and deep learning advances show potential for ECG-based CVD detection, but a critical analysis of model performance, dataset features, preprocessing, and explainable artificial intelligence was needed.

Do machine learning and deep learning models provide high diagnostic performance for ECG-based cardiovascular disease prediction?

Population

75 included studies evaluating ECG-based ML and DL models for CVD prediction

Comparison

ECG-based ML/DL models with or without explainable AI vs benchmark or standard validation

Design

Systematic review

Key result

Machine learning and deep learning models for ECG-based cardiovascular disease prediction demonstrated high diagnostic performance, with accuracies commonly ranging from 90% to 99% across 75 studies.

Authors

SPSabit Ahamed PreantoDaffodil International UniversityMBMd. Hasan Imam BijoyDaffodil International UniversityTPTapon PaulDaffodil International University

Discussion

Loading...

Member takes

Overview

May accelerate AI-ECG adoption in clinics; extends systematic evidence on model performance while highlighting validation gaps.

Key Points

  • To critically review recent machine learning, deep learning, and explainable artificial intelligence models for ECG-based cardiovascular disease detection and identify key challenges in clinical translation.
  • Systematically evaluated 75 studies focusing on ECG-based artificial intelligence applications across conditions including arrhythmia, myocardial infarction, heart failure, and atrial fibrillation.
  • Analyzed model architectures (e.g., CNN, LSTM, Transformers, SVM, RF), benchmark datasets (e.g., MIT-BIH, PTB-XL, UK Biobank), preprocessing techniques, and explainability methods.
  • Across 75 included studies, reported diagnostic accuracy ranged from approximately 90% to 99%, sensitivity from 81% to 99%, specificity from 84% to 99%, AUROC up to 0.98–0.99, and F1-scores exceeded 0.90 on benchmark datasets.
  • Predominant architectures included CNN, LSTM, Transformer, SVM, RF, and hybrid CNN-LSTM networks applied to raw signals, wearable ECG data, and ECG images.
  • Translational barriers persist due to reliance on internal validation, class imbalance, and a notable absence of prospective clinical and external validation.

Study Design

Type

Systematic Review (n=75)

Structured PICO

Do machine learning and deep learning models provide high diagnostic performance for ECG-based cardiovascular disease prediction?

P
Population
75 peer-reviewed studies published between 2021 and 2025 evaluating machine learning and deep learning models for ECG-based cardiovascular disease prediction.
I
Intervention
Machine learning (ML) and deep learning (DL) models (including CNN, LSTM, Transformer, SVM, RF, and hybrid CNN-LSTM architectures) for ECG analysis
O
Outcome
Diagnostic performance (accuracy, sensitivity, specificity, AUROC, and F1-score)surrogate

While machine learning and deep learning models demonstrate excellent internal diagnostic performance for ECG-based cardiovascular disease prediction, their clinical translation is currently limited by a lack of external and prospective validation.

Limitations

  • Class imbalance
  • Data quality variation
  • Limited generalizability
  • Model interpretability issues
  • Limited availability of external and prospective clinical validation
  • Potential publication bias from excluding lower-ranked journals and conference proceedings
  • class imbalance
  • data quality variation
  • limited generalizability
  • model interpretability
  • limited availability of external and prospective clinical validation
  • ethical, legal, and social issues (data bias, privacy, transparency, clinical trust)

Cite This Study

Preanto et al. (2026) conducted a systematic review in Cardiovascular disease (n=75). Machine learning and deep learning models was evaluated on Diagnostic performance (accuracy, sensitivity, specificity, AUROC). Machine learning and deep learning models for ECG-based cardiovascular disease prediction demonstrated high diagnostic performance, with accuracies commonly ranging from 90% to 99% across 75 studies.

synapsesocial.com/papers/6a93e9743affd7f5e23e6b2dhttps://doi.org/10.1007/s44163-026-02070-w
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Systematic Review of Artificial Intelligence and Electrocardiography for Cardiovascular Disease Diagnosis2025
  2. 2Systematic Review of Artificial Intelligence and Electrocardiography for Cardiovascular Disease Diagnosis2025 · 13 citations
  3. 3A systematic survey of artificial intelligence methods for ECG-based cardiovascular disease prediction2026
  4. 4Machine Learning Approaches for Automated Diagnosis of Cardiovascular Diseases: A Review of Electrocardiogram Data Applications2024 · 1 citations
  5. 5Advances in artificial intelligence techniques for diagnosis of cardiac diseases2026