What is the clinical evidence from this study?

Study design: Systematic Review. Population: Cardiovascular disease (n=83). Intervention: Artificial intelligence methods (machine learning and deep learning).

What does this research mean for the field?

While artificial intelligence models demonstrate strong performance in controlled settings for ECG-based cardiovascular disease prediction, most literature suffers from critical methodological limitations including a frequent lack of external validation, high risk of bias, and limited demographic diversity. Novelty: ClaimNovelty.SYNTHESIS. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

To systematically evaluate artificial intelligence methods for ECG-based cardiovascular disease prediction across data handling, modeling, interpretability, and validation practices.

September 28, 2026Discover Artificial IntelligenceOpen Access

Deep learning ECG studies reveal significant methodological gaps, with ~72% lacking external validation.

View Full Paper

Ask AI

Bookmark

Population

83 peer-reviewed studies applying artificial intelligence methods to ECG-based cardiovascular disease…

Design

Systematic_review

Key result

While deep learning approaches for ECG analysis demonstrate impressive performance, significant methodological gaps remain, with only 38.6% of studies showing low risk of bias and 72% lacking external validation.

Authors

TMThanuja M SPSoumyashree M. Panchal PSPrasanna Lakshmi G S

Discussion

Member takes

Overview

Clinicians should await externally validated, low-bias AI-ECG tools; challenges overoptimistic internal-performance claims across the literature.

Key Points

To systematically evaluate artificial intelligence methods for ECG-based cardiovascular disease prediction across data handling, modeling, interpretability, and validation practices.
Followed PRISMA guidelines to search PubMed, IEEE Xplore, Scopus, Web of Science, and major databases for studies published between January 2020 and November 2025.
Screened 1,247 records and identified 83 studies for comprehensive assessment using a customized tool evaluating data preprocessing, model architecture, and clinical readiness.
Deep learning models, particularly convolutional neural networks and hybrid architectures, achieved strong performance in controlled settings, but only 38.6% of studies showed a low risk of bias.
External validation was absent in 72% of studies, with 78% of all training data derived from North America or Europe and minimal inclusion of pediatric cohorts.
Explainable AI methods were implemented in only 52% of recent evaluations, and demographic fairness assessments remained rare.

Study Design

Type

Systematic Review (n=83)

Structured PICO

Population

A systematic review of 83 peer-reviewed studies published between 2020 and 2025 evaluating artificial intelligence methods for ECG-based cardiovascular disease prediction.

Intervention

Artificial intelligence methods (including deep learning, convolutional neural networks, and hybrid architectures)

Outcome

Methodological practices including data handling, modeling approaches, interpretability techniques, and evaluation strategies

This systematic review highlights significant methodological gaps in AI-ECG research, including a lack of external validation and dataset diversity, providing a roadmap for developing reliable and fair AI systems for cardiovascular care.

Limitations

Only 38.6% of studies showed low risk of bias.
External validation was lacking in 72% of cases.
Dataset diversity remains problematic, with 78% of data originating from North America or Europe and minimal representation of pediatric populations.
Explainable AI methods appear in only 52% of recent studies, and demographic fairness assessments remain rare.
Lack of external validation in 72% of cases
Dataset diversity is problematic (78% of data from North America or Europe)
Minimal representation of pediatric populations
Explainable AI methods appear in only 52% of recent studies
Demographic fairness assessments remain rare

Cite This Study

M et al. (2026) conducted a systematic review in Cardiovascular disease (n=83). Artificial intelligence methods (machine learning and deep learning) was evaluated. While deep learning approaches for ECG analysis demonstrate impressive performance, significant methodological gaps remain, with only 38.6% of studies showing low risk of bias and 72% lacking external validation.

synapsesocial.com/papers/6ab9b5067822ec8fc3d8f52c https://doi.org/10.1007/s44163-026-02170-7

View Full Paper

Ask AI

Bookmark

September 28, 2026Discover Artificial IntelligenceOpen Access

Deep learning ECG studies reveal significant methodological gaps, with ~72% lacking external validation.

View Full Paper

Ask AI

Bookmark

Population

83 peer-reviewed studies applying artificial intelligence methods to ECG-based cardiovascular disease…

Design

Systematic_review

Key result

Authors

TMThanuja M SPSoumyashree M. Panchal PSPrasanna Lakshmi G S

Discussion

Member takes

Overview

Clinicians should await externally validated, low-bias AI-ECG tools; challenges overoptimistic internal-performance claims across the literature.

Key Points

To systematically evaluate artificial intelligence methods for ECG-based cardiovascular disease prediction across data handling, modeling, interpretability, and validation practices.
Followed PRISMA guidelines to search PubMed, IEEE Xplore, Scopus, Web of Science, and major databases for studies published between January 2020 and November 2025.
Screened 1,247 records and identified 83 studies for comprehensive assessment using a customized tool evaluating data preprocessing, model architecture, and clinical readiness.
Deep learning models, particularly convolutional neural networks and hybrid architectures, achieved strong performance in controlled settings, but only 38.6% of studies showed a low risk of bias.
External validation was absent in 72% of studies, with 78% of all training data derived from North America or Europe and minimal inclusion of pediatric cohorts.
Explainable AI methods were implemented in only 52% of recent evaluations, and demographic fairness assessments remained rare.

Study Design

Type

Systematic Review (n=83)

Structured PICO

Population

A systematic review of 83 peer-reviewed studies published between 2020 and 2025 evaluating artificial intelligence methods for ECG-based cardiovascular disease prediction.

Intervention

Artificial intelligence methods (including deep learning, convolutional neural networks, and hybrid architectures)

Outcome

Methodological practices including data handling, modeling approaches, interpretability techniques, and evaluation strategies

Limitations

Only 38.6% of studies showed low risk of bias.
External validation was lacking in 72% of cases.
Dataset diversity remains problematic, with 78% of data originating from North America or Europe and minimal representation of pediatric populations.
Explainable AI methods appear in only 52% of recent studies, and demographic fairness assessments remain rare.
Lack of external validation in 72% of cases
Dataset diversity is problematic (78% of data from North America or Europe)
Minimal representation of pediatric populations
Explainable AI methods appear in only 52% of recent studies
Demographic fairness assessments remain rare

Cite This Study

synapsesocial.com/papers/6ab9b5067822ec8fc3d8f52c https://doi.org/10.1007/s44163-026-02170-7

Also Consider

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

View Full Paper

Ask AI

Bookmark