Interpretable Multimodal Machine Learning for Early Stroke Risk‬ ‭ Prediction in Atrial Fibrillation Patients

Anand Hadapad; Jayanthi M

Interpretable Multimodal Machine Learning for Early Stroke Risk‬ ‭ Prediction in Atrial Fibrillation Patients

July 2025

Vol-11, Issue-4

Paper ID: 27084

ISSN: 2395-4396

Downloads: 0

Abstract & Details

Research Area

Computer Science and Engineering

Keywords

Atrial Fibrillation (AF) Stroke Risk Prediction Electrocardiogram (ECG) RR Intervals Heart Rate Variability (HRV) Machine Learning (ML) XGBoost time-domain features multimodal data analysis signal processing artificial neural network (ANN) SMOTE Feature Importance Clinical Decision Support.

Abstract

Atrial Fibrillation (AF) is a prevalent cardiac arrhythmia that significantly increases the risk of stroke and other cardiovascular complications. Early and accurate detection of AF is critical for timely intervention, yet traditional methods rely heavily on manual analysis of electrocardiogram (ECG) signals, which is time-consuming and prone to error. This study presents an interpretable, multimodal machine learning approach for detecting AF episodes and enabling early stroke risk prediction based on heart rate variability (HRV) features derived from ECG signals.The dataset, sourced from the Erasmus Medical Centre, includes ECG recordings of 804 post-operative patients. RR intervals were extracted using a semi-automatic pipeline and converted into 30-second segments with binary AF labels. A comprehensive feature engineering pipeline generated 30 HRV features spanning the time domain, frequency domain, geometrical, non-linear (CSI and CVI), and Poincaré plot representations. Various machine learning models—including Logistic Regression, Naïve Bayes, KNN-DTW, Random Forest, Artificial Neural Networks (ANN), and XGBoost—were trained using these features.Among all models, the XGBoost classifier demonstrated the highest performance with 99% accuracy and 0.99 recall on the test set. Interpretability was achieved through feature importance analysis, which revealed that time-domain features such as std_hr, pnni_20, and median_nni were the most predictive.This study highlights the potential of interpretable machine learning models for automatic AF detection and sets the stage for future work involving stroke risk prediction by integrating multimodal data, such as clinical parameters and patient demographics.

License

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Author Information

#	Name	Institute / Affiliation
1	Anand Hadapad	CMR University
2	Jayanthi M	Assistant Professor CMR University

How to Cite

Use the following formats to cite this article in your research.

APA Style

Hadapad, Anand & M, Jayanthi (2025). Interpretable Multimodal Machine Learning for Early Stroke Risk‬ ‭ Prediction in Atrial Fibrillation Patients. International Journal of Advance Research and Innovative Ideas In Education, 11(4), 581-588.

MLA Style

Hadapad, Anand, and Jayanthi M. "Interpretable Multimodal Machine Learning for Early Stroke Risk‬ ‭ Prediction in Atrial Fibrillation Patients." International Journal of Advance Research and Innovative Ideas In Education, vol. 11, no. 4, 2025, pp. 581-588.

IEEE Style

Anand Hadapad and Jayanthi M, "Interpretable Multimodal Machine Learning for Early Stroke Risk‬ ‭ Prediction in Atrial Fibrillation Patients," International Journal of Advance Research and Innovative Ideas In Education, vol. 11, no. 4, pp. 581-588, 2025.

Vancouver Style

Hadapad Anand, M Jayanthi. Interpretable Multimodal Machine Learning for Early Stroke Risk‬ ‭ Prediction in Atrial Fibrillation Patients. International Journal of Advance Research and Innovative Ideas In Education. 2025;11(4):581-588.

Harvard Style

Hadapad, Anand & M, Jayanthi (2025) 'Interpretable Multimodal Machine Learning for Early Stroke Risk‬ ‭ Prediction in Atrial Fibrillation Patients', International Journal of Advance Research and Innovative Ideas In Education, 11(4), pp. 581-588.

Chicago Style

Hadapad, Anand and Jayanthi M. "Interpretable Multimodal Machine Learning for Early Stroke Risk‬ ‭ Prediction in Atrial Fibrillation Patients." International Journal of Advance Research and Innovative Ideas In Education 11, no. 4 (2025): 581-588.

Turabian Style

Hadapad, Anand and Jayanthi M. "Interpretable Multimodal Machine Learning for Early Stroke Risk‬ ‭ Prediction in Atrial Fibrillation Patients." International Journal of Advance Research and Innovative Ideas In Education 11, no. 4 (2025): 581-588.