Abstract

Early and accurate prediction of cardiovascular disease (CVD) is vital for improving clinical outcomes. Traditional diagnostic tools often struggle to capture the complex temporal and morphological characteristics embedded in electrocardiogram (ECG) and phonocardiogram (PCG) signals. This study introduces a multi-algorithm optimization framework that leverages hybrid neural network architectures for CVD detection. Nonlinear autoregressive models with exogenous Inputs (NARX) are employed to capture temporal dependencies in ECG signals. In contrast, Neural network fitting (NNF) and neural network pattern recognition (NNPR) models are utilized for classifying PCG signals. The training performance of these models is evaluated using different optimization algorithms, including Levenberg marquardt (LM), Scaled conjugate gradient (SCG), and Bayesian regularization (BR). The NARX-BR techniques outperformed all others, achieving an R-squared is 96.8% and 98.2% classification accuracy. ECG signals yielded higher predictive performance than PCG, attributed to lower noise and clearer temporal features. BR demonstrated superior generalizations, while SCG offered faster convergence with competitive results. The proposed hybrid neural network framework, combined with robust preprocessing and optimization strategies, significantly enhances the accuracy of CVD detection. These findings underscore the potential of advanced neural models for non-invasive, early-stage cardiovascular diagnosis and continuous monitoring.

Keywords

Machine learning, NARX, Neural network, PCG, ECG, Feature extraction, Classification models, Cardiovascular disease diagnosis, Signal processing.

References

[1] Münzel T, Sørensen M, Hahad O, Nieuwenhuijsen M, Daiber A. The contribution of the exposome to the burden of cardiovascular disease. Nature Reviews Cardiology. 2023; 20(10):651-69.

[2] Tique MR, Araque O, Sanchez-echeverri LA. Technological advances in the diagnosis of cardiovascular disease: a public health strategy. International Journal of Environmental Research and Public Health. 2024; 21(8):1-15.

[3] Aabdalla ID, Vasumathi D. Wavelet scattering transform for ECG cardiovascular disease classification. International Journal of Artificial Intelligence and Applications. 2024; 15(1):99-111.

[4] Aabdalla ID, Vasumathi D. A comprehensive systematic review for cardiovascular disease using machine learning techniques. International Journal of Artificial Intelligence and Applications. 2024; 15(1):1-23.

[5] Ahmad S, Asghar MZ, Alotaibi FM, Alotaibi YD. Retracted article: diagnosis of cardiovascular disease using deep learning technique. Soft Computing. 2023; 27(13):8971-90.

[6] Nikhar S, Karandikar AM. Prediction of heart disease using machine learning algorithms. International Journal of Advanced Engineering, Management and Science. 2016; 2(6):617-21.

[7] Benouar S, Kedir-talha M, Seoane F. Time-series NARX feedback neural network for forecasting impedance cardiography ICG missing points: a predictive model. Frontiers in Physiology. 2023; 14:1181745.

[8] Hangaragi S, Neelima N, Jegdic K, Nagarwal A. Integrated fusion approach for multi-class heart disease classification through ECG and PCG signals with deep hybrid neural networks. Scientific Reports. 2025; 15(1):8129.

[9] Bahrami S, Shamsollahi MB, Ansari V. Investigating causal relationships between cardiovascular signals using effective connectivity assessment measures. In 30th national and 8th international Iranian conference on biomedical engineering 2023 (pp. 1-7). IEEE.

[10] Sumon MS, Islam MS, Rahman MS, Hossain MS, Khandakar A, Hasan A, et al. CardioTabNet: a novel hybrid transformer model for heart disease prediction using tabular medical data. arXiv preprint arXiv:2503.17664. 2025.

[11] Alromema W, Alduweib E, Abduh Z. Heart sound classification using the nonlinear dynamic feature approach along with conventional classifiers. Engineering, Technology & Applied Science Research. 2023; 13(3):10808-13.

[12] Elola A, Aramendi E, Irusta U, Picón A, Alonso E, Owens P, et al. Deep neural networks for ECG-based pulse detection during out-of-hospital cardiac arrest. Entropy. 2019; 21(3):305.

[13] Paul B, Karn B. Heart disease prediction using scaled conjugate gradient backpropagation of artificial neural network. Soft Computing. 2023; 27(10):6687-702.

[14] Fayaz SA, Zaman M, Kaul S, Bakshi WJ. Optimizing cardiovascular disease prediction: a synergistic approach of grey wolf levenberg model and neural networks. Journal of Information Systems Engineering & Business Intelligence. 2023; 9(2):119-35.

[15] Jyothi P, Pradeepini G. Heart disease detection system based on ECG and PCG signals with the aid of GKVDLNN classifier. Multimedia Tools and Applications. 2024; 83(10):30587-612.

[16] Ghorashi S, Rehman K, Riaz A, Alkahtani HK, Samak AH, Cherrez-ojeda I, et al. Leveraging regression analysis to predict overlapping symptoms of cardiovascular diseases. IEEE Access. 2023; 11:60254-66.

[17] Mohamed N, Awasthi MA, Kulkarni N, Thota S, Singh M, Dhole SV. Intelligent systems and applications in engineering. International Journal of Intelligent Systems and Applications in Engineering. 2022:212-6.

[18] Nia PS, Hesar HD. Abnormal heart sound detection using time-frequency analysis and machine learning techniques. Biomedical Signal Processing and Control. 2024; 90:105899.

[19] Rao GM, Ramesh D, Sharma V, Sinha A, Hassan MM, Gandomi AH. AttGRU-HMSI: enhancing heart disease diagnosis using hybrid deep learning approach. Scientific Reports. 2024; 14(1):7833.

Cite this article