Hybrid CNN-LSTM Architecture: Early Heart Disease Prediction from Electronic Health Records
Keywords:
Attention mechanism, Cardiovascular disease prediction, CNN-LSTM hybrid model, Deep learning, Electronic health records, Heart diseaseAbstract
Cardiovascular diseases (CVDs) remain the leading cause of mortality in India, accounting for 20% of global heart attack deaths and contributing to over 32,000 fatalities annually. Early prediction of heart disease is crucial for timely intervention and reducing premature mortality. This study proposes a novel hybrid deep learning architecture combining convolutional neural networks (CNN) and long short-term memory (LSTM) networks with a multi-visit attention mechanism to predict heart disease from electronic health records (EHR). Unlike existing models that process single-visit data, this architecture analyzes sequential patient visits (up to 5 temporal sequences) to capture disease progression patterns. The proposed model extracts spatial features using CNN layers, processes temporal dependencies through LSTM networks, and employs a self-attention mechanism to prioritize critical clinical features across multiple patient visits. By validating this approach on the UCI heart disease dataset (303 patients) and a pilot subset of 50 Indian patient records, 91.2% accuracy in preliminary trials was achieved. Full validation on 500 patients is planned for extended work, incorporating region-specific risk factors, including diabetes prevalence, hypertension, lifestyle patterns, and healthcare accessibility.
References
D. H. Barlow, “A prolegomenon to clinical psychology: Two 40-year odysseys,” in The Oxford Handbook of Clinical Psychology, D. H. Barlow, Ed. Oxford, U.K.: Oxford University Press, pp. 3–20.
L. Loitongbam and W. R. Surin, “The rising burden of cardiovascular disease and thrombosis in India: An epidemiological review,” Cureus, vol. 16, no. 11, p. e73786, Nov. 2024.
D. Ghosh, “India accounts for 20% of global heart attack deaths, says new Indian report,” NDTV, Feb. 21, 2026.
M. Shannawaz, I. Rathi, N. Shah, S. Saeed, A. Chandra, and H. Singh, “Prevalence of CVD among Indian adult population: Systematic review and meta-analysis,” International Journal of Environmental Research and Public Health, vol. 22, no. 4, pp. 539–539, Apr. 2025.
E. Maini, B. Venkateswarlu, B. Maini, and D. Marwaha, “Machine learning–based heart disease prediction system for Indian population: An exploratory study done in South India,” Medical Journal Armed Forces India, vol. 77, no. 3, pp. 302–311, Jan. 2021.
S. Raghavendra et al., “DLMNN-based heart disease prediction with PD-SS optimization algorithm,” Intelligent Automation & Soft Computing, vol. 35, no. 2, pp. 1353–1368, 2023.
V. K. Sudha and D. Kumar, “Hybrid CNN and LSTM network for heart disease prediction,” SN Computer Science, vol. 4, no. 2, Jan. 2023.
N. Mishra and G. B. Shivaji, “Advancing heart disease forecasting: Integrating CNN-LSTM models with feature enhancement for improved predictive accuracy,” Journal of Angiotherapy, vol. 8, no. 9, pp. 1–6, Sep. 2024.
E. Haque, M. Paul, and F. Tohidi, “Predicting heart failure with attention learning techniques utilizing cardiovascular data,” arXiv preprint, Jul. 2024.
A. U. Rahman, Y. Alsenani, A. Zafar, K. Ullah, K. M. Rabie, and T. Shongwe, “Enhancing heart disease prediction using a self-attention-based transformer model,” Scientific Reports, vol. 14, no. 1, Jan. 2024.
S. Alyounis, A. Khandoker, C. Stefanini, and L. J. Hadjileontiadis, “A hybrid CNN-LSTM model for heart failure detection using raw ECG signals,” in Proc. Computing in Cardiology Conf. (CinC), vol. 51, Dec. 2024, pp. 1–4.
Md. T. Hayat et al., “A hybrid convolutional neural network–long short-term memory (CNN–LSTM)–attention model architecture for precise medical image analysis and disease diagnosis,” Diagnostics, vol. 15, no. 21, p. 2673, Oct. 2025.
A. Panday, “Heart attack risk & prediction dataset in India,” Kaggle Dataset, 2025.
S. Ulianova, “Cardiovascular disease dataset,” Kaggle Dataset, 2019.
H. Kaur, “A hybrid CNN-transformer model for enhanced heart disease prediction,” in Proc. 7th Int. Conf. Signal Processing, Computing and Control (ISPCC), Solan, India, 2025, pp. 991–996.
