Cardiol Res

Cardiology Research, ISSN 1923-2829 print, 1923-2837 online, Open Access

Article copyright, the authors; Journal compilation copyright, Cardiol Res and Elmer Press Inc

Journal website https://cr.elmerpub.com

Review

Volume 17, Number 5, October 2026, pages 406-415

Prediction of Risk of Cardiovascular Events in Patients With Stable Angina Using Artificial Intelligence: A Systematic Review

↓ Figure 1. The PRISMA flow diagram for this systematic review. PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

↓ **Table 1.** Inclusion and Exclusion Criteria
Category	Inclusion criteria	Exclusion criteria
MI: myocardial infarction; EHR: electronic health record; ECG: electrocardiogram; MACE: major adverse cardiovascular events.
Population	Adult patients (≥ 18 years) diagnosed with stable angina or chronic coronary syndrome	Patients with acute coronary syndrome, unstable angina, post-MI only cohorts, pediatric populations
Intervention/exposure	Studies applying artificial intelligence, machine learning, or deep learning models	Studies using only traditional statistical risk scores without AI/ML components
Outcomes	Prediction of cardiovascular events (e.g., MACE, MI, mortality, heart failure, stroke, revascularization)	Studies not reporting cardiovascular outcomes or focused only on diagnostic accuracy without outcome prediction
Study design	Observational studies (retrospective or prospective), cohort studies, registry-based studies	Case reports, editorials, letters, commentaries, conference abstracts, posters, dissertations
Data type	Use of clinical data, ECG, imaging, EHR, laboratory data, or multimodal datasets	Animal studies, simulation-only studies, non-clinical datasets
Publication type	Peer-reviewed journal articles	Preprints without peer review, short communications, reviews
Time frame	Published within the last 10 years	Publications which were prior to the specified time window
Language	Articles published in English	Non-English publications

↓ Table 1. Inclusion and Exclusion Criteria

Category

Inclusion criteria

Exclusion criteria

MI: myocardial infarction; EHR: electronic health record; ECG: electrocardiogram; MACE: major adverse cardiovascular events.

Population

Adult patients (≥ 18 years) diagnosed with stable angina or chronic coronary syndrome

Patients with acute coronary syndrome, unstable angina, post-MI only cohorts, pediatric populations

Intervention/exposure

Studies applying artificial intelligence, machine learning, or deep learning models

Studies using only traditional statistical risk scores without AI/ML components

Outcomes

Prediction of cardiovascular events (e.g., MACE, MI, mortality, heart failure, stroke, revascularization)

Studies not reporting cardiovascular outcomes or focused only on diagnostic accuracy without outcome prediction

Study design

Observational studies (retrospective or prospective), cohort studies, registry-based studies

Case reports, editorials, letters, commentaries, conference abstracts, posters, dissertations

Data type

Use of clinical data, ECG, imaging, EHR, laboratory data, or multimodal datasets

Animal studies, simulation-only studies, non-clinical datasets

Publication type

Peer-reviewed journal articles

Preprints without peer review, short communications, reviews

Time frame

Published within the last 10 years

Publications which were prior to the specified time window

Language

Articles published in English

Non-English publications

↓ Table 2. Summarized Literature for the Systematic Review

Reference

Target

Variable

Architecture

Pre-processing

Dataset

Selection Criteria and Outcome

Result

ACS: acute coronary syndrome; AHA: American Heart Association; BMI: body mass index; CAD: coronary artery disease; CCTA: coronary computed tomography angiography; CHD: coronary heart disease; CI: confidence interval; CNN: convolutional neural network; CV: cardiovascular; CVE: cardiovascular events; D-F: Diamond–Forrester; DM: diabetes mellitus; ECG: electrocardiogram; EHR: electronic health records; ExtCAD: extensive coronary artery disease; FHx: family history; HF: heart failure; HTN: hypertension; MI: myocardial infarction; ObstCAD: obstructive coronary artery disease; QCG: quantitative electrocardiography; QC: quality control; UA: unstable angina; vars: variables; WHO: World Health Organization.

[17]

AI-ECG detection of obstructive CAD in stable angina

Age, sex, BMI, symptom type, HTN, DM, dyslipidemia, smoking, stroke, FHx + QCG scores + ECG deep features

ResNet-CNN (16 layers, SE blocks, nonlocal); task layers (Adam + focal loss; temp scaling). XGBoost on extracted vectors (80/20; 5-fold CV)

Signal extraction, feature vectors, probability calibration, CV for tuning

SNUBH, 2018–2020, n = 723

Outpatients with stable angina symptoms undergoing invasive angiography. Excluded: asymptomatic, prior CAD/revascularization, and emergency ACS cases.

AUROC: 0.827

Outcome: any obstructive CAD

[18]

Obstructive + extensive CAD in suspected stable angina

Clinical risk factors + QCG scores

Modified ResNet encoder

Patient-level split (90/10), bootstrap CI (2000)

SNUBH, 2011–2019, n = 21,866

Outpatients evaluated via invasive angiography/CCTA for chest pain. External validation: independent multi-cohort sample. Excluded: emergency ACS admissions.

AUC: ObstCAD 0.781; severe 0.780; ExtCAD 0.689

Outcome: obstructive CAD

[13]

Obstructive CAD and assess prognosis in suspected stable angina

Age, sex, angina typicality, dyspnoa; risk factors (DM, dyslipidemia, FHx); Charlson index

Statistical models: logistic regression (basic + clinical); Cox regression for prognosis

Symptom classification (typical/atypical/non-cardiac), D-F pre-test probability, registry linkage, 30-day blanking period

Single-center, Denmark, n = 3,903

Consecutive single-center outpatient referrals. Baseline status: verified free of baseline CAD and heart failure at entry.

AUC: 0.88

Outcome: obstructive CAD

Composite outcome (death, MI, UA, HF, stroke)

[19]

12-month CVE in stable angina + CHD

Physical activity, Antiplatelet use, Traditional Chinese medicine, Gensini score, SAQ Exertional capacity, SAQ Anginal stability

ML models: XGBoost, LightGBM, RF, AdaBoost, MLP, SVM, KNN, GNB) + logistic regression,

Prospective cohort, random split 7:3, 10-fold CV, missing MICE imputation, SHAP for interpretability

Single-center, n = 827

Prospective cohort with established WHO/AHA-defined stable coronary disease. Excluded: advanced systemic illness, heart failure, or cognitive barriers.

AUC: train 1.00, validation 0.98, test 0.98 (95% CI 0.97–1.00)

Best = LightGBM; also kept multivariable logistic regression (nomogram + online calculator)

Outcome: composite CVE: nonfatal MI, revascularization, all-cause death, readmission (angina/HF/malignant arrhythmia), stroke

[20]

ACS risk in stable CHD

EHR vars → QC → 34 clinical vars (labs + diagnoses + utilization + sociodemographics)

T2G-Former + MLP: build feature-relation graph → transformer embeddings → MLP outputs ACS baselines: CatBoost, LightGBM, XGBoost, RF, SVM, FT-Transformer

Convert heterogeneous variables, imputation, patient-level longitudinal mean, MICE (checked vs k-NN), multicollinearity check 5-fold stratified CV, repeated across 12 seeds

Raw: 268,876 CHD records from 165 institutions, n = 12,336

Outpatients with stable angina symptoms undergoing invasive angiography.

The F1 scores were 0.666, 0.757, and 0.886 at 6, 12, and 24 months, respectively.

Outcome: progression from stable CHD → ACS within 6/12/24 months