A GRU-based Model for Early Stroke Prediction Using Surface Electromyography Signals

Bob Chile-Agada; Laud Charles Ochei; Fubara Egbono

doi:doi:10.11648/j.ijiis.20261502.12

Research Article |

| Peer-Reviewed

A GRU-based Model for Early Stroke Prediction Using Surface Electromyography Signals

Bob Chile-Agada^*

, Laud Charles Ochei

, Fubara Egbono

Published in International Journal of Intelligent Information Systems (Volume 15, Issue 2)

Received: 15 August 2026 Accepted: 31 August 2026 Published: 20 September 2026

Views: Downloads:

Download PDF

Share This Article

Twitter
Linked In
Facebook

Abstract

Stroke is a leading cause of mortality and long-term disability globally, necessitating early detection for timely intervention. This paper presents a comprehensive study on the development, training, and evaluation of a Gated Recurrent Unit (GRU) model for early stroke prediction using Surface Electromyography (sEMG) signals. The proposed model leverages the temporal modeling capabilities of recurrent neural networks to analyze sequential EMG data from 8 channels, intended to capture neuromuscular signal characteristics associated with stroke-induced motor impairment. The model architecture includes a GRU layer with 64 hidden units, followed by a fully connected layer with ReLU activation and a final softmax output layer for binary classification. We implement and compare two variants: a GRU-only model and a CNN-GRU hybrid. Training incorporates gradient clipping by norm and learning rate schedulers (step decay and exponential decay) to address vanishing/exploding gradient problems and optimize convergence. The model is trained and evaluated using the MUSED-I sEMG dataset, comprising 11 healthy subjects and 2 stroke patients, with a total of 11,146 windowed samples. Group K-Fold cross-validation (K=5) is used, though with only 5 resulting groups (3 healthy-file, 2 stroke-patient), some folds are single-class by construction rather than providing full subject-independent generalization evidence (see Section 3.7). The GRU-only model achieved a higher mean cross-validation accuracy (74.9%, SD 7.3%) than the CNN-GRU model (55.3%) and a Logistic Regression baseline (47.8%), though a paired comparison did not reach statistical significance (p = 0.125; see Section 5.1). An Integrated Gradients analysis on a single real test example showed differential channel importance concentrated on Channels 4 and 5; averaging across the test set is needed before generalizing this finding. The final model's Brier score (0.242) is reported for completeness, though it was computed on a single evaluation split containing no stroke-labeled test examples (see Section 4.3.2), so it reflects calibration on the negative class only; a supplementary, non-group-independent re-evaluation on a class-balanced split (Section 4.3.3) confirms the architecture can achieve genuine discrimination (ROC-AUC = 0.839, PR-AUC = 0.752) once both classes are present in the test set, at the cost of subject independence. Despite challenges related to dataset limitations and class imbalance, the results demonstrate the feasibility of GRU-based approaches for non-invasive, cost-effective stroke screening, providing a preliminary feasibility signal to guide future model refinement in a larger, prospectively validated cohort.

Published in	International Journal of Intelligent Information Systems (Volume 15, Issue 2)
DOI	10.11648/j.ijiis.20261502.12
Page(s)	34-52
Creative Commons	This is an Open Access article, distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium or format, provided the original work is properly cited.
Copyright	Copyright © The Author(s), 2026. Published by Science Publishing Group

Keywords

Stroke Prediction, Surface Electromyography, Gated Recurrent Unit, Deep Learning, Early Detection, Signal Processing, Recurrent Neural Networks, Group K-Fold Cross-Validation

References

[1]	Adam, S. Y., Yousif, A., & Bashir, M. B. (2016). Classification of ischemic stroke using machine learning algorithms. International Journal of Computer Application, 149(10), 26-31. https://doi.org/10.5120/ijca2016911607
[2]	Alhakeem, A., Chaurasia, B., & Khan, M. M. (2025). Revolutionizing stroke prediction: A systematic review of AI-powered wearable technologies for early detection of stroke. Neurosurgical Review, 48(1), 458. https://doi.org/10.1007/s10143-025-03629-4
[3]	Aviles, M., Alvarez-Alvarado, J. M., Robles-Ocampo, J.-B., Sevilla-Camacho, P. Y., & Rodríguez-Reséndiz, J. (2024). Optimizing RNNs for EMG signal classification: A novel strategy using Grey Wolf Optimization. Bioengineering, 11(1), 77. https://doi.org/10.3390/bioengineering11010077
[4]	Bass, L., Clements, P., & Kazman, R. (2012). Software architecture in practice (3rd ed.). Addison-Wesley.
[5]	Cahuantzi, R., Chen, X., & Güttel, S. (2021). A comparison of LSTM and GRU networks for learning symbolic sequences https://doi.org/10.48550/arXiv.2107.02248
[6]	Çelik, Y., & Can, Ü. (2026). Surface EMG-based hand gesture recognition using a hybrid multistream deep learning architecture. Sensors, 26(7), 2281. https://doi.org/10.3390/s26072281
[7]	Cho, K., van Merrienboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014). Learning phrase representations using RNN encoder-decoder for statistical machine translation. arXiv preprint arXiv:1406.1078. https://doi.org/10.3115/v1/D14-1179
[8]	Choi, Y.-A., Park, S.-J., Jun, J.-A., Pyo, C.-S., Cho, K.-H., Lee, H.-S., & Yu, J.-H. (2021). Deep learning-based stroke disease prediction system using real-time bio signals. Sensors, 21(13), 4269. https://doi.org/10.3390/s21134269
[9]	Cuocolo, R., Cipullo, M. B., Stanzione, A., et al. (2019). Machine learning applications in prostate cancer magnetic resonance imaging. European Radiology Experimental, 3(1), 35. https://doi.org/10.1186/s41747-019-0109-2
[10]	Guntari, E. W., Djamal, E. C., Nugraha, F., & Liem, S. L. L. (2020). Classification of post-stroke EEG signal using genetic algorithm and recurrent neural networks. Proceedings of the 2020 7th International Conference on Electrical Engineering, Computer Sciences and Informatics, 156-161.
[11]	Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735-1780. https://doi.org/10.1162/neco.1997.9.8.1735
[12]	Ioffe, S., & Szegedy, C. (2015). Batch normalization: Accelerating deep network training by reducing internal covariate shift. Proceedings of the 32nd International Conference on Machine Learning (ICML 2015), 448-456. https://doi.org/10.48550/arXiv.1502.03167
[13]	Jha, A., Aicher, J. K., Gazzara, M. R., Singh, D., & Barash, Y. (2020). Enhanced Integrated Gradients: Improving interpretability of deep learning models using splicing codes as a case study. Genome Biology, 21, 149. https://doi.org/10.1186/s13059-020-02055-7
[14]	Joloudari, J. H., Marefat, A., Nematollahi, M. A., Oyelere, S. S., & Hussain, S. (2023). Effective class-imbalance learning based on SMOTE and convolutional neural networks. Applied Sciences, 13(6), 4006. https://doi.org/10.3390/app13064006
[15]	Kingma, D. P., & Ba, J. (2015). Adam: A method for stochastic optimization. Proceedings of the 3rd International Conference on Learning Representations (ICLR 2015). https://doi.org/10.48550/arXiv.1412.6980
[16]	Kourou, K., Exarchos, T. P., Exarchos, K. P., Karamouzis, M. V., & Fotiadis, D. I. (2015). Machine learning applications in cancer prognosis and prediction. Computational and Structural Biotechnology Journal, 13, 8-17. https://doi.org/10.1016/j.csbj.2014.11.005
[17]	Lakkshmanan, A., Suji, A., Priyanka, N., Anand, D. B., Nagaraju, K., & Naidu, U. U. (2023). A novel machine learning based stroke prediction system using magnetic resonance image and adaptive new fuzzy inference. Intelligent Systems and Applications in Engineering, 12(10), 576-585.
[18]	Loshchilov, I., & Hutter, F. (2019). Decoupled weight decay regularization. Proceedings of the 7th International Conference on Learning Representations (ICLR 2019). https://doi.org/10.48550/arXiv.1711.05101
[19]	Marwat, M. (2023). MUSED-I: Surface Electromyography (sEMG) dataset [Data set]. Kaggle. https://www.kaggle.com/datasets/mustafamarwat/mused-i-semg-dataset-for-stroke-rehabilitation
[20]	Schuster, M., & Paliwal, K. K. (1997). Bidirectional recurrent neural networks. IEEE Transactions on Signal Processing, 45(11), 2673-2681. https://doi.org/10.1109/78.650093
[21]	Shinde, A., Shete, V., & Mehendale, N. (2026). Optimized signal acquisition and advanced AI for robust 1D EMG classification: A comparative study of machine learning, deep learning, and reinforcement learning. Bioengineering, 13(4), 463. https://doi.org/10.3390/bioengineering13040463
[22]	World Health Organization. (2019). Global Health Estimates: Life expectancy and leading causes of death and disability. Geneva: WHO Press.
[23]	Xia, X., Yue, W., Chao, B., Li, M., Cao, L., Wang, L., Shen, Y., & Li, X. (2019). Prevalence and risk factors of stroke in the elderly in Northern China: Data from the National Stroke Screening Survey. Journal of Neurology, 266, 1449-1458. https://doi.org/10.1007/s00415-019-09281-5
[24]	Yu, J., Park, S., Kwon, S. H., Ho, C. M. B., Pyo, C. S., & Lee, H. (2020). AI-based stroke disease prediction system using real-time electromyography signals. Applied Sciences, 10, 6791. https://doi.org/10.3390/app10196791
[25]	Yuan, J., Wu, F., & Wu, H. (2023). Multivariate time-series classification using memory and attention for long and short-term dependence. Applied Intelligence, 53(24), 29677-29692. https://doi.org/10.1007/s10489-023-05079-1

Cite This Article

Plain Text BibTeX RIS

APA Style

Chile-Agada, B., Ochei, L. C., Egbono, F. (2026). A GRU-based Model for Early Stroke Prediction Using Surface Electromyography Signals. International Journal of Intelligent Information Systems, 15(2), 34-52. https://doi.org/10.11648/j.ijiis.20261502.12

Copy | Download

ACS Style

Chile-Agada, B.; Ochei, L. C.; Egbono, F. A GRU-based Model for Early Stroke Prediction Using Surface Electromyography Signals. Int. J. Intell. Inf. Syst. 2026, 15(2), 34-52. doi: 10.11648/j.ijiis.20261502.12

Copy | Download

AMA Style

Chile-Agada B, Ochei LC, Egbono F. A GRU-based Model for Early Stroke Prediction Using Surface Electromyography Signals. Int J Intell Inf Syst. 2026;15(2):34-52. doi: 10.11648/j.ijiis.20261502.12

Copy | Download

@article{10.11648/j.ijiis.20261502.12,
  author = {Bob Chile-Agada and Laud Charles Ochei and Fubara Egbono},
  title = {A GRU-based Model for Early Stroke Prediction Using Surface Electromyography Signals},
  journal = {International Journal of Intelligent Information Systems},
  volume = {15},
  number = {2},
  pages = {34-52},
  doi = {10.11648/j.ijiis.20261502.12},
  url = {https://doi.org/10.11648/j.ijiis.20261502.12},
  eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.ijiis.20261502.12},
  abstract = {Stroke is a leading cause of mortality and long-term disability globally, necessitating early detection for timely intervention. This paper presents a comprehensive study on the development, training, and evaluation of a Gated Recurrent Unit (GRU) model for early stroke prediction using Surface Electromyography (sEMG) signals. The proposed model leverages the temporal modeling capabilities of recurrent neural networks to analyze sequential EMG data from 8 channels, intended to capture neuromuscular signal characteristics associated with stroke-induced motor impairment. The model architecture includes a GRU layer with 64 hidden units, followed by a fully connected layer with ReLU activation and a final softmax output layer for binary classification. We implement and compare two variants: a GRU-only model and a CNN-GRU hybrid. Training incorporates gradient clipping by norm and learning rate schedulers (step decay and exponential decay) to address vanishing/exploding gradient problems and optimize convergence. The model is trained and evaluated using the MUSED-I sEMG dataset, comprising 11 healthy subjects and 2 stroke patients, with a total of 11,146 windowed samples. Group K-Fold cross-validation (K=5) is used, though with only 5 resulting groups (3 healthy-file, 2 stroke-patient), some folds are single-class by construction rather than providing full subject-independent generalization evidence (see Section 3.7). The GRU-only model achieved a higher mean cross-validation accuracy (74.9%, SD 7.3%) than the CNN-GRU model (55.3%) and a Logistic Regression baseline (47.8%), though a paired comparison did not reach statistical significance (p = 0.125; see Section 5.1). An Integrated Gradients analysis on a single real test example showed differential channel importance concentrated on Channels 4 and 5; averaging across the test set is needed before generalizing this finding. The final model's Brier score (0.242) is reported for completeness, though it was computed on a single evaluation split containing no stroke-labeled test examples (see Section 4.3.2), so it reflects calibration on the negative class only; a supplementary, non-group-independent re-evaluation on a class-balanced split (Section 4.3.3) confirms the architecture can achieve genuine discrimination (ROC-AUC = 0.839, PR-AUC = 0.752) once both classes are present in the test set, at the cost of subject independence. Despite challenges related to dataset limitations and class imbalance, the results demonstrate the feasibility of GRU-based approaches for non-invasive, cost-effective stroke screening, providing a preliminary feasibility signal to guide future model refinement in a larger, prospectively validated cohort.},
 year = {2026}
}

Copy | Download

TY  - JOUR
T1  - A GRU-based Model for Early Stroke Prediction Using Surface Electromyography Signals
AU  - Bob Chile-Agada
AU  - Laud Charles Ochei
AU  - Fubara Egbono
Y1  - 2026/09/20
PY  - 2026
N1  - https://doi.org/10.11648/j.ijiis.20261502.12
DO  - 10.11648/j.ijiis.20261502.12
T2  - International Journal of Intelligent Information Systems
JF  - International Journal of Intelligent Information Systems
JO  - International Journal of Intelligent Information Systems
SP  - 34
EP  - 52
PB  - Science Publishing Group
SN  - 2328-7683
UR  - https://doi.org/10.11648/j.ijiis.20261502.12
AB  - Stroke is a leading cause of mortality and long-term disability globally, necessitating early detection for timely intervention. This paper presents a comprehensive study on the development, training, and evaluation of a Gated Recurrent Unit (GRU) model for early stroke prediction using Surface Electromyography (sEMG) signals. The proposed model leverages the temporal modeling capabilities of recurrent neural networks to analyze sequential EMG data from 8 channels, intended to capture neuromuscular signal characteristics associated with stroke-induced motor impairment. The model architecture includes a GRU layer with 64 hidden units, followed by a fully connected layer with ReLU activation and a final softmax output layer for binary classification. We implement and compare two variants: a GRU-only model and a CNN-GRU hybrid. Training incorporates gradient clipping by norm and learning rate schedulers (step decay and exponential decay) to address vanishing/exploding gradient problems and optimize convergence. The model is trained and evaluated using the MUSED-I sEMG dataset, comprising 11 healthy subjects and 2 stroke patients, with a total of 11,146 windowed samples. Group K-Fold cross-validation (K=5) is used, though with only 5 resulting groups (3 healthy-file, 2 stroke-patient), some folds are single-class by construction rather than providing full subject-independent generalization evidence (see Section 3.7). The GRU-only model achieved a higher mean cross-validation accuracy (74.9%, SD 7.3%) than the CNN-GRU model (55.3%) and a Logistic Regression baseline (47.8%), though a paired comparison did not reach statistical significance (p = 0.125; see Section 5.1). An Integrated Gradients analysis on a single real test example showed differential channel importance concentrated on Channels 4 and 5; averaging across the test set is needed before generalizing this finding. The final model's Brier score (0.242) is reported for completeness, though it was computed on a single evaluation split containing no stroke-labeled test examples (see Section 4.3.2), so it reflects calibration on the negative class only; a supplementary, non-group-independent re-evaluation on a class-balanced split (Section 4.3.3) confirms the architecture can achieve genuine discrimination (ROC-AUC = 0.839, PR-AUC = 0.752) once both classes are present in the test set, at the cost of subject independence. Despite challenges related to dataset limitations and class imbalance, the results demonstrate the feasibility of GRU-based approaches for non-invasive, cost-effective stroke screening, providing a preliminary feasibility signal to guide future model refinement in a larger, prospectively validated cohort.
VL  - 15
IS  - 2
ER  -

Copy | Download

Author Information

Bob Chile-Agada

Department of Computer Science, University of Port Harcourt, Choba, Nigeria

Contact Email

http://orcid.org/0000-0002-0338-1908
Laud Charles Ochei

Department of Computer Science, University of Port Harcourt, Choba, Nigeria

Contact Email

http://orcid.org/0000-0003-4148-1085
Fubara Egbono

Department of Computer Science, University of Port Harcourt, Choba, Nigeria

Contact Email

Download PDF

Submit an Article

Sections

Plain Text BibTeX RIS

APA Style

Chile-Agada, B., Ochei, L. C., Egbono, F. (2026). A GRU-based Model for Early Stroke Prediction Using Surface Electromyography Signals. International Journal of Intelligent Information Systems, 15(2), 34-52. https://doi.org/10.11648/j.ijiis.20261502.12

Copy | Download

ACS Style

Chile-Agada, B.; Ochei, L. C.; Egbono, F. A GRU-based Model for Early Stroke Prediction Using Surface Electromyography Signals. Int. J. Intell. Inf. Syst. 2026, 15(2), 34-52. doi: 10.11648/j.ijiis.20261502.12

Copy | Download

AMA Style

Chile-Agada B, Ochei LC, Egbono F. A GRU-based Model for Early Stroke Prediction Using Surface Electromyography Signals. Int J Intell Inf Syst. 2026;15(2):34-52. doi: 10.11648/j.ijiis.20261502.12

Copy | Download

@article{10.11648/j.ijiis.20261502.12,
  author = {Bob Chile-Agada and Laud Charles Ochei and Fubara Egbono},
  title = {A GRU-based Model for Early Stroke Prediction Using Surface Electromyography Signals},
  journal = {International Journal of Intelligent Information Systems},
  volume = {15},
  number = {2},
  pages = {34-52},
  doi = {10.11648/j.ijiis.20261502.12},
  url = {https://doi.org/10.11648/j.ijiis.20261502.12},
  eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.ijiis.20261502.12},
  abstract = {Stroke is a leading cause of mortality and long-term disability globally, necessitating early detection for timely intervention. This paper presents a comprehensive study on the development, training, and evaluation of a Gated Recurrent Unit (GRU) model for early stroke prediction using Surface Electromyography (sEMG) signals. The proposed model leverages the temporal modeling capabilities of recurrent neural networks to analyze sequential EMG data from 8 channels, intended to capture neuromuscular signal characteristics associated with stroke-induced motor impairment. The model architecture includes a GRU layer with 64 hidden units, followed by a fully connected layer with ReLU activation and a final softmax output layer for binary classification. We implement and compare two variants: a GRU-only model and a CNN-GRU hybrid. Training incorporates gradient clipping by norm and learning rate schedulers (step decay and exponential decay) to address vanishing/exploding gradient problems and optimize convergence. The model is trained and evaluated using the MUSED-I sEMG dataset, comprising 11 healthy subjects and 2 stroke patients, with a total of 11,146 windowed samples. Group K-Fold cross-validation (K=5) is used, though with only 5 resulting groups (3 healthy-file, 2 stroke-patient), some folds are single-class by construction rather than providing full subject-independent generalization evidence (see Section 3.7). The GRU-only model achieved a higher mean cross-validation accuracy (74.9%, SD 7.3%) than the CNN-GRU model (55.3%) and a Logistic Regression baseline (47.8%), though a paired comparison did not reach statistical significance (p = 0.125; see Section 5.1). An Integrated Gradients analysis on a single real test example showed differential channel importance concentrated on Channels 4 and 5; averaging across the test set is needed before generalizing this finding. The final model's Brier score (0.242) is reported for completeness, though it was computed on a single evaluation split containing no stroke-labeled test examples (see Section 4.3.2), so it reflects calibration on the negative class only; a supplementary, non-group-independent re-evaluation on a class-balanced split (Section 4.3.3) confirms the architecture can achieve genuine discrimination (ROC-AUC = 0.839, PR-AUC = 0.752) once both classes are present in the test set, at the cost of subject independence. Despite challenges related to dataset limitations and class imbalance, the results demonstrate the feasibility of GRU-based approaches for non-invasive, cost-effective stroke screening, providing a preliminary feasibility signal to guide future model refinement in a larger, prospectively validated cohort.},
 year = {2026}
}

Copy | Download

TY  - JOUR
T1  - A GRU-based Model for Early Stroke Prediction Using Surface Electromyography Signals
AU  - Bob Chile-Agada
AU  - Laud Charles Ochei
AU  - Fubara Egbono
Y1  - 2026/09/20
PY  - 2026
N1  - https://doi.org/10.11648/j.ijiis.20261502.12
DO  - 10.11648/j.ijiis.20261502.12
T2  - International Journal of Intelligent Information Systems
JF  - International Journal of Intelligent Information Systems
JO  - International Journal of Intelligent Information Systems
SP  - 34
EP  - 52
PB  - Science Publishing Group
SN  - 2328-7683
UR  - https://doi.org/10.11648/j.ijiis.20261502.12
AB  - Stroke is a leading cause of mortality and long-term disability globally, necessitating early detection for timely intervention. This paper presents a comprehensive study on the development, training, and evaluation of a Gated Recurrent Unit (GRU) model for early stroke prediction using Surface Electromyography (sEMG) signals. The proposed model leverages the temporal modeling capabilities of recurrent neural networks to analyze sequential EMG data from 8 channels, intended to capture neuromuscular signal characteristics associated with stroke-induced motor impairment. The model architecture includes a GRU layer with 64 hidden units, followed by a fully connected layer with ReLU activation and a final softmax output layer for binary classification. We implement and compare two variants: a GRU-only model and a CNN-GRU hybrid. Training incorporates gradient clipping by norm and learning rate schedulers (step decay and exponential decay) to address vanishing/exploding gradient problems and optimize convergence. The model is trained and evaluated using the MUSED-I sEMG dataset, comprising 11 healthy subjects and 2 stroke patients, with a total of 11,146 windowed samples. Group K-Fold cross-validation (K=5) is used, though with only 5 resulting groups (3 healthy-file, 2 stroke-patient), some folds are single-class by construction rather than providing full subject-independent generalization evidence (see Section 3.7). The GRU-only model achieved a higher mean cross-validation accuracy (74.9%, SD 7.3%) than the CNN-GRU model (55.3%) and a Logistic Regression baseline (47.8%), though a paired comparison did not reach statistical significance (p = 0.125; see Section 5.1). An Integrated Gradients analysis on a single real test example showed differential channel importance concentrated on Channels 4 and 5; averaging across the test set is needed before generalizing this finding. The final model's Brier score (0.242) is reported for completeness, though it was computed on a single evaluation split containing no stroke-labeled test examples (see Section 4.3.2), so it reflects calibration on the negative class only; a supplementary, non-group-independent re-evaluation on a class-balanced split (Section 4.3.3) confirms the architecture can achieve genuine discrimination (ROC-AUC = 0.839, PR-AUC = 0.752) once both classes are present in the test set, at the cost of subject independence. Despite challenges related to dataset limitations and class imbalance, the results demonstrate the feasibility of GRU-based approaches for non-invasive, cost-effective stroke screening, providing a preliminary feasibility signal to guide future model refinement in a larger, prospectively validated cohort.
VL  - 15
IS  - 2
ER  -

Copy | Download