Explainable AI-Based Intelligent Closed-Loop Drug Delivery System for Mean Arterial Blood Pressure (MABP) Using Controlled Drug Administration
Abstract
Mean Arterial Blood Pressure (MABP) regulation using vasoactive drugs remains challenging in intensive care units due to nonlinear patient dynamics, inter-patient variability, and limitations of manual titration protocols. This study introduces an ensemble closed-loop controller that integrates Proportional-Integral-Derivative (PID), Fuzzy Logic, Model Predictive Control (MPC), and Reinforcement Learning (RL) algorithms, augmented by SHapley Additive exPlanations (SHAP)-based Explainable AI (XAI) for real-time arterial signal processing and interpretable decision-making. The system employs a physiologically validated nonlinear patient model simulating MABP dynamics under realistic Gaussian noise conditions. Reinforcement Learning demonstrates superior setpoint tracking and settling performance, while MPC provides strong foresight-based optimization; PID and Fuzzy Logic offer robust classical and expert-knowledge-driven baselines. An ensemble fusion strategy weights contributions from each algorithm to leverage complementary strengths, addressing individual limitations like computational intensity or data requirements. XAI integration delivers real-time feature attributions highlighting trajectory lag and error as dominant drivers building clinician trust for regulatory approval. PySpark scalability supports multi-patient validation, with a structured pathway from synthetic cohorts to randomized controlled trials, promising enhanced hemodynamic stability and reduced nursing workload in clinical deployment.
References
K. E. Kwok, S. L. Shah, B. A. Finegan, and G. K. Kwong, “Experiences with experimental clinical evaluation of a computerized drug delivery system for regulation of mean arterial blood pressure,” in Proceedings of the American Control Conference, 1999, vol. 2, pp. 1264–1268Y.
Gao and M. J. Er, “Adaptive modeling and control of drug delivery systems using generalized fuzzy neural networks,” in Advances in Intelligent Systems, Springer, 2004, pp. 327–346.
M. J. Er and Y. Zhang, “Adaptive modeling and intelligent control of a sodium nitroprusside delivery system,” in Artificial Neural Networks for Drug Design, Delivery and Disposition. Academic Press, 2016, pp. 333–354.
G. K. Kwong, K. E. Kwok, B. A. Finegan, and S. L. Shah, “Clinical evaluation of long-range adaptive control for mean arterial blood pressure regulation,” in Proceedings of the American Control Conference, Seattle, WA, USA, Jun. 1995, vol. 1, pp. 786–790.
N. Malagutti, A. Dehghani, and R. A. Kennedy, “A robust multiple-model adaptive control system for automatic delivery of a vasoactive drug,” Proceedings of the International Association of Science and Technology for Development, 2012.
H. Ying and L. C. Sheppard, “Fuzzy control of mean arterial pressure in postsurgical patients,” in Proceedings of the Third Workshop on Neural Networks: Academic/Industrial/NASA/Defense, Feb. 1992, pp. 521–524.
B. Singh and S. Urooj, “Adaptive parameter estimation-based drug delivery system for blood pressure regulation,” in Advances in Intelligent Systems and Computing, Springer, 2018, pp. 465–472.
R. Zhang, Z. Li, X. Pan, Z. Ma, Y. Dai, A. Mohammadzadeh, and C. Zhang, “Adaptive drug delivery to control mean arterial blood pressure by reinforcement fuzzy Q-learning,” IEEE Sensors Journal, vol. 24, no. 19, pp. 30968–30977, Aug. 2024.
P. Asfar et al., “High versus low blood-pressure target in patients with septic shock,” The New England Journal of Medicine, vol. 370, no. 17, pp. 1583–1593, Apr. 2014.
P. Tahmasebi and A. Hezarkhani, “A hybrid neural networks–fuzzy logic–genetic algorithm for grade estimation,” Computers & Geosciences, vol. 42, pp. 18–27, May 2012.
R. Chandramouli, “Proportional resonance and fractional-order proportional integral derivative-based closed-loop drug infusion for the regulation of mean arterial pressure in critical care patients: A modeling and simulation study,” Asian Journal of Pharmaceutics, vol. 14, no. 1, 2020
S. A. Nirmala, R. Muthu, and B. V. Abirami, “Drug infusion control for mean arterial pressure regulation of critical care patients,” in Proceedings of the IEEE International Conference on TENCON, 2013, pp. 1–4
S. A. Nirmala, R. Muthu, and B. V. Abirami, “Model predictive control of drug infusion system for mean arterial pressure regulation of critical care patients,” Research Journal of Applied Sciences, Engineering and Technology, vol. 7, no. 21, pp. 4601–4605, Jun. 2014.
X. Hong, W. Ayadi, K. A. Alattas, A. Mohammadzadeh, M. Salimi, and C. Zhang, “Adaptive average arterial pressure control by multi-agent on-policy reinforcement learning,” Scientific Reports, vol. 15, no. 1, p. 679, Jan. 2025.
V. T. Mai, K. A. Alattas, Y. Bouteraa, E. Ghaderpour, and A. Mohammadzadeh, “Personalized blood pressure control by machine learning for remote patient monitoring,” IEEE Access, vol. 12, pp. 83994–84004, 2024.
H. Ying, M. McEachern, D. W. Eddleman, and L. C. Sheppard, “Adaptive fuzzy control of mean arterial pressure in critically ill patients,” in Proceedings of the American Control Conference, San Francisco, CA, USA, Jun. 1993, pp. 1845–1849.
