Development and validation of a deep neural network for predicting coronary heart disease in hypertensive patients using 24-hour ambulatory blood pressure monitoring: a retrospective study
The developed deep neural network model enables early identification of high-risk CHD patients with hypertension through interpretable, routinely available clinical variables through interpretable, routinely available clinical variables.
Abstract
Background Coronary heart disease (CHD) remains a leading cause of morbidity and mortality worldwide. Early identification of high-risk hypertensive patients is crucial for preventing cardiovascular events. While traditional risk scores rely on static clinical measurements, 24-h ambulatory blood pressure monitoring (ABPM)-derived time in target range (TTR) captures dynamic blood pressure control patterns that may improve risk stratification. Machine learning methods, particularly deep neural networks, offer an enhanced capability to model complex non-linear relationships in high-dimensional clinical data, compared with conventional statistical approaches. Methods This single-center retrospective cohort study included 1,026 patients admitted between January 2023 and December 2024, with 718 patients allocated to model development and 308 to internal validation. A deep neural network model with three hidden layers was developed and compared against eight conventional machine learning algorithms (logistic regression, naïve Bayes, k-nearest neighbors, random forest, support vector machine, XGBoost, LightGBM, and CatBoost). Thirty-two variables spanning demographics, clinical data, laboratory results, echocardiographic measures, and blood pressure indices were evaluated. Continuous variables were discretized into quartile-based categories to enhance clinical interpretability. Feature selection employed a two-step process combining the Boruta algorithm and least absolute shrinkage and selection operator (LASSO) regression, with variance inflation factor analysis confirming the absence of collinearity. Model selection prioritized balanced performance across discrimination (AUC), calibration (Brier score), and clinical utility (decision curve analysis) in the independent validation cohort. Interpretability was evaluated using SHAP (SHapley Additive exPlanations) values. Results The deep neural network model achieved optimal balanced performance with an AUC of 0.822 (95% CI: 0.793–0.850) in the training cohort and 0.796 (95% CI: 0.749–0.846) in the validation cohort, accompanied by the lowest Brier score (0.172), indicating superior calibration. Nine predictors were retained: diabetes mellitus, mean systolic blood pressure, time in target range of systolic blood pressure, left atrial diameter, left ventricular end-systolic diameter, left ventricular ejection fraction, use of antihypertensive medications, calcium channel blockers, and β-blockers. SHAP analysis identified TTR and blood pressure control parameters as the primary drivers of model predictions. Conclusion The developed deep neural network model enables early identification of high-risk CHD patients with hypertension through interpretable, routinely available clinical variables. Prospective multicenter external validation is warranted to confirm its generalizability across diverse populations and clinical settings.
Hypertension is one of the most important modifiable risk factors for Cardiovascular Disease (CVD), yet identifying which hypertensive patients are at higher risk remains challenging in clinical practice. This study developed and evaluated three machine-learning models: logistic regression, random forest, and Gradient Boosting for CVD risk prediction in a cohort of 23,543 hypertensive patients drawn from a 70,000 patient cardiovascular dataset. After preprocessing, feature engineering, SMOTE-based class balancing, and hyperparameter tuning via randomized search, model performance was assessed on a held-out test set and validated using 5-fold stratified cross-validation with SMOTE correctly nested inside each fold to avoid data leakage. On the test set, tuned Gradient Boosting model achieved the highest accuracy (78.59%) and AUC-ROC (0.6681), outperforming Logistic Regression (0.6633) and Random Forest (0.6508). cross-validation provided a slightly different perspective: Logistic Regression’s mean AUC-ROC (0.6628) edged out Gradient Boosting (0,6609) and Random Forest (0.6383), SHAP analysis on the Gradient Boosting model identified systolic blood pressure, age, and height as the strongest predictors, with height rivaling systolic blood pressure and surpassing BMI a notable difference from Random Forest’s feature importance ranking. Lifestyle factors (smoking, alcohol, physical activity) contributed minimally. These findings highlight blood pressure and body size measures as the dominant clinical signals in this dataset, while demonstrating the potential of an explainable machine-learning model based on routinely collected clinical data to support cardiovascular risk stratification and clinical decision-making in hypertensive patients, despite their moderate discriminative performance.
C. M. Anyanwu, J. C. Onyianta, Ogechi Gift Onyedi et al.· Nature Journal of Emerging S...· 0 citations
Cardiovascular Diseases (CVDs) continue to be one of the leading causes of deaths in the world, claiming some 17.9 million lives every year. This burden is higher in Pakistan because of "Asian Indian Phenotype" which makes them vulnerable to early coronary artery disease. The commonly used traditional risk prediction models, including the Framingham Risk Score, have been developed in Western populations and are poorly predictive in South Asian populations. This study aims to fill this important gap by designing, implementing and comparative evaluation of six supervised machine learning algorithms for early detection of cardiovascular disease using a locally collected clinical dataset of 411 patient records with 13 independent clinical attributes. The models tested are Logistic Regression, K Nearest Neighbor, Support Vector Machine, Random Forest, Gradient Boosting and XGBoost. A rigorous gender-based mean imputation and Z-score normalization was done and split in 80/20 ratio. Empirical results show that the Random Forest classifier has Area under the Curve (AUC) of 0.9842, accuracy of 95.2%, precision of 96.0% and recall of 96.0%. The model was then exported and used to create a browser-based, predictive application that could be embedded in an interactive dashboard for real-time cardiovascular risk without the need for a server. These results confirm the effectiveness of ensemble learning approaches for medical diagnostics and highlight the potential of implementing ML-based screening tools in the limited resource healthcare environment in Pakistan.
Awais Khursheed, Soban Ahmed, Sibghat Ullah et al.· International Journal of Inn...· 0 citations
Abstract Background Patients undergoing dialysis are at an elevated risk of cardiovascular events. This study aimed to develop machine learning (ML) prediction models to identify risk factors for major adverse cardiovascular events (MACE) in dialysis patients. Materials and Methods This retrospective study included 203 patients undergoing dialysis with a median age of 45.0 years and 64.0% male. The participants were divided into training and test sets in a 7:3 ratio. LASSO regression selected characteristic variables from patients’general information, laboratory tests, and echocardiographic parameters (including global longitudinal strain [GLS]). Eight ML models were constructed,and SHAP analysis evaluated feature importance. Results The incidence of MACE (including myocardial infarction, unstable angina, heart failure, and cardiovascular death) in dialysis patients was 38.92%. The average follow-up period was 18 months. LASSO regression identified eight feature variables. Among the ML models, AdaBoost demonstrated superior performance, with an AUC of 0.883 (95% CI: 0.830–0.937), accuracy of 0.804, sensitivity of 0.864 and specificity of 0.762 in the training set, and an AUC of 0.809 (95% CI: 0.706–0.912), accuracy of 0.750, sensitivity of 0.90 and specificity of 0.675 in the test set. The SHAP analysis identified N-terminal pro-brain natriuretic peptide (NT-proBNP) level, estimated glomerular filtration rate (eGFR), GLS and age as the four most important features for predicting MACE in patients undergoing dialysis (mean absolute SHAP values: 0.199, 0.176, 0.096 and 0.091, respectively). Conclusion Elevated NT-proBNP, advanced age, reduced eGFR and impaired GLS were independently associated with an increased risk of MACE in patients undergoing dialysis.
Mei Jin, Zikang Lin, Lingxiang Ma et al.· Annals medicus· 0 citations
Hypertension, commonly known as high blood pressure, is a major risk factor for cardiovascular diseases and premature mortality worldwide. Early detection and prevention are critical in reducing its health impact. This study explores the application of machine learning (ML) techniques to predict the likelihood of hypertension in individuals using clinical and demographic data. A variety of supervised learning algorithms, including Logistic Regression, Random Forest, Support Vector Machines, and Gradient Boosting, were evaluated for their predictive performance [1]. The dataset was preprocessed through feature selection, normalization, and handling of missing values to improve model accuracy.[2] Performance metrics such as accuracy, precision, recall, F1-score, and AUC-ROC were used to assess the models [4]. The results demonstrate that ML models can effectively identify individuals at high risk of hypertension, offering a valuable tool for early intervention and personalized healthcare [5]. This approach underscores the potential of artificial intelligence in supporting public health efforts and enhancing clinical decision-making.
Key words: Logistic Regression, Random Forest, Support Vector Machines, and Gradient Boosting.
G. Vamsi, K. Bhargavi· International Scientific Jou...· 0 citations
Male sex was a statistically significant independent predictor of heart disease after controlling for other clinical variables and the findings support sex-specific screening and preventive strategies for high-cholesterol male patients and demonstrate the value of interpretable machine learning models for clinical decision support.
Taiwo Samson Adeyemo· GSC Advanced Research and Re...· 0 citations
BACKGROUND
Coronary artery disease (CAD) is the leading cause of death globally and a major contributor to hospital readmission. This study aimed to predict 30-day mortality in patients hospitalized with acute and chronic CAD using a structured machine learning approach with data from multiple centers.
METHODS
We conducted a retrospective cohort study using patient data from the Taipei Medical University Clinical Research Database (TMUCRD). Multiple machine learning algorithms were employed to develop predictive models for 30-day mortality. Model performance was evaluated using a stratified fivefold cross-validation approach. Key performance metrics included the area under the curve (AUC), accuracy, sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), and F1 score.
RESULTS
A total of 23,267 patients (mean age 64.9 years) were included, with 1215 deaths overall (5.2%): 570 (3.7%) in the internal cohort (n = 15,510) and 645 (8.3%) in the external validation cohort (n = 7757, Shuang Ho Hospital). XGBoost achieved the best performance for the overall and acute CAD cohorts (AUROC 0.845 and 0.820, respectively), while logistic regression performed best for chronic CAD (AUROC 0.766). Key predictive features included the Charlson Comorbidity Index, hemoglobin level, emergency room admission status, age, and creatinine level.
CONCLUSION
The use of a structured machine learning approach to predict 30-day mortality in patients with acute and chronic CAD demonstrated promising discriminative performance, providing valuable insights that could enhance personalized care and inform clinical decisions.
Septi Melisa, P. Phan, Sheng-Hsuan Chien et al.· International Journal of Car...· 0 citations