A Hybrid Clinical Decision Support System for Heart Attack Risk Prediction with SHAP-Based Explainability and Evidence-Based Clinical Interpretation

Abstract

Heart disease is a major cause of death globally, therefore leading to the need for reliable and interpretable predictive models to support early diagnosis. Current machine learning techniques for predicting heart disease predominantly concentrate on boosting accuracy, while overlooking interpretability and clinical actionability. Here we present a combined clinical decision support system that leverages machine learning and explainable AI to deliver predictive performance with meaningful clinical explanations. Our approach integrates an XGBoost machine learning model (optimised for non-ECG datasets) with a second ECG-informed feature stream, which is then meta-learnt to better account for variations in predictions and increase model robustness. To retain the crucial aspect of explainability in medical scenarios, SHAP (SHapley Additive exPlanations) explanations are used and translated into reliable evidence-based clinical interpretations through well-accepted medical thresholds and guidelines. Two-layer explanation interface to the clinicians and patients also characterize the framework and enhance explainability and usability. It is experimentally and tested using the UCI Heart Disease data set where the system has good predictive performance (with a predicted area of about 0.911) and predictive probability. Additionally, the stability test of the noisy and missing data situation warrants the stability of the proposed system. The findings suggest that the methodology that relies on the combination of hybrid models with explainable and clinically reasoning may make AI-based solutions more utilized in clinical environments in predicting the risks of cardiovascular complications.

Introduction

The text describes a multimodal deep learning system designed to detect fake news using different types of data such as text, images, videos, and audio. It works by collecting data from multiple sources, preprocessing it (cleaning text, resizing images, extracting video frames, and processing audio), and extracting meaningful features using NLP and CNN-based techniques.

The system then uses deep learning models like CNN, RNN/LSTM, and BERT to classify the content. A fusion mechanism combines results from all media types to improve accuracy, producing a final prediction indicating whether the content is real, fake, or manipulated, along with confidence scores.

The architecture includes key stages: input, preprocessing, feature extraction, deep learning processing, fusion, classification, and output, making it suitable for real-time and scalable applications like media monitoring and fact-checking.

Conclusion

In the present paper, a hybrid clinical decision support mechanism to predict the risk of heart attack based on a combination of machine learning and explainable artificial intelligence with clinical interpretation is presented. The technique combines XGBoost prediction, ECG-inspired feature stream/stacking fusion to enhance the accuracy and stability of models. The experimentation of the system depicts high predictive accuracy, precise modeling of probability and resilience to missing and spurious data. SHAP interpretation and evidence-based clinical reasoning helps the system to offer both clinical interpretability and practical information to both clinicians and patients. The system surpasses the weak points of existing methods by creating a system that integrates predictive performance with clinical interpretability. With such orthogonal properties as interpretability, robustness and knowledge, the system fosters the development of actionable AI-based systems in healthcare. Future research is directed at one of the priorities to expand the proposed system to accommodate bigger datasets, raw sensor data streams, and test its feasibility in clinical practice.

References

[1] C. M. Bhatt, P. Patel, T. Ghetia, and P. L. Mazzeo, “Effective heart disease prediction using machine learning techniques,” Algorithms, vol. 16, no. 2, p. 88, 2023. [2] T. Roopa and G. D. Ramanjinappa, “Heart disease predictive modeling with XGBoost and SMOTE-driven class imbalance mitigation,” Engineering, Technology & Applied Science Research, vol. 15, no. 6, pp. 29914–29918, 2025. [3] J. Yang and J. Guan, “A heart disease prediction model based on feature optimization and SMOTE-XGBoost algorithm,” Information, vol. 13, no. 10, p. 475, 2022. [4] S. Dhandapani, H. Somasundaram, and T. Angamuthu, “Hybrid deep learning framework for heart disease prediction using ECG signal images,” Scientific Reports, vol. 15, 2025. [5] P. Archana and S. V. Shashikala, “A hybrid deep learning heart disease prediction framework utilizing multi-modal medical imaging and novel feature fusion techniques,” Engineering, Technology & Applied Science Research, vol. 15, 2025. [6] D. M. AlSekait and M. Zakariah, \"Using convolutional neural networks with late fusion to predict heart disease,\" Scientific Reports, vol. 15, p. 41260, 2025. DOI: 10.1038/s41598-025-25100-6 [7] P. Shah, M. Shukla, N. H. Dholakia, and H. Gupta, “Predicting cardiovascular risk with hybrid ensemble learning and explainable AI,” Scientific Reports, vol. 15, pp. 1–21, 2025. [8] S. M. Ganie, P. K. D. Pramanik, and Z. Zhao, “Ensemble learning with explainable AI for improved heart disease prediction based on multiple datasets,” Scientific Reports, vol. 15, p. 13912, 2025. [9] C. Li et al., “Improving cardiovascular risk prediction through machine learning modelling of irregularly repeated electronic health records,” European Heart Journal—Digital Health, vol. 5, no. 1, pp. 30–40, 2024. [10] H. El-Sofany et al., “A proposed technique for predicting heart disease using machine learning algorithms and an explainable AI method,” Scientific Reports, vol. 14, p. 23277, 2024. [11] C. J. Ejiyi et al., “XAI-HD: An explainable artificial intelligence framework for heart disease detection,” Artificial Intelligence Review, 2025. [12] Z. Kucukakcali and I. B. Cicek, \"Beyond black boxes: Interpretable AI with explainable neural networks for acute myocardial infarction using common hematological parameters,\"Medicina, vol. 61, no. 9, p. 1552 2025. DOI: 10.3390/medicina61091552 [13] A. Adekoya, F. Saeed, W. Ghaban, and S. N. Qasem, \"Ensemble learning approach with explainable AI for improved heart disease prediction,\" Frontiers in Pharmacology, vol. 16, p. 1654681, 2025. DOI: 10.3389/fphar.2025.1654681 [14] A. B. Yazdipour, P. Mehdizadeh, M. Arabfard, M. Yaghoubi, M. M. Farahani, and M. Amiri-Ara, \"Development and implementation of explainable AI-based machine learning models for predicting hospital stay and treatment costs in cardiovascular patients,\" Scientific Reports, vol. 16, p. 3154, 2026. DOI: 10.1038/s41598-025-33005-7 [15] R. Kumar, S. Garg, R. Kaur, M. G. M. Johar, S. Singh, S. V. Menon, P. Kumar, A. M. Hadi, S. A. Hasson, and J. Lozanovic, \"A comprehensive review of machine learning for heart disease prediction: challenges, trends, ethical considerations, and future directions,\" Frontiers in Artificial Intelligence, vol. 8, p. 1583459, 2025. DOI: 10.3389/frai.2025.1583459 [16] S. M. Lundberg and S. I. Lee, “A unified approach to interpreting model predictions,” in Advances in Neural Information Processing Systems (NeurIPS), 2017. [17] T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” in Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, pp. 785–794. [18] T. Akiba, S. Sano, T. Yanase, T. Ohta, and M. Koyama, \"Optuna: A next-generation hyperparameter optimization framework,\" in Proc. ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2019, pp. 2623–2631. [19] A. Janosi, W. Steinbrunn, M. Pfisterer, and R. Detrano, \"Heart Disease Dataset,\" UCI Machine Learning Repository, 1989. [Online]. Available: https://archive.ics.uci.edu/dataset/45/heart+disease [20] L. Grinsztajn, E. Oyallon, and G. Varoquaux, “Why tree-based models still outperform deep learning on tabular data,” in NeurIPS, 2022. [21] R. Shwartz-Ziv and A. Armon, “Tabular data: Deep learning is not all you need,” Information Fusion, vol. 81, pp. 84–90, 2022.

Copyright

Copyright © 2026 Nikhil Yadav, Mayank Soni, Sachin Kumar, Kartik Poddar, Anurag Kumar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.