DL-PRWarnNeuro: A Transformer-Based Deep Learning Framework for Multimodal Patient Deterioration Prediction in Neurological Intensive Care
Abstract
Monitoring vital signs is extremely important in neurological intensive care, as even minor changes can indicate a significant decline in the patient's condition. The purpose of this work is to develop DL-PRWarnNeuro, a deep learning model intended to provide early, accurate alerts in neurological intensive care settings by leveraging patient signal data from the MIMIC-IV dataset. A convolutional layer is incorporated into the model's architecture to extract local patterns. This is followed by transformer-based attentive layers, designed to focus on crucial signal fluctuations. During training, Focal Loss is used to address class imbalance between typical, stable states and rare, critical events. The performance of DL-PRWarnNeuro is tested against baseline models, including LSTM classifiers and standard threshold-based monitoring. Among the most important indicators are the early detection rate, the False Alarm Rate (FAR), and the Area Under the Receiver Operating Characteristics curve (AUROC). Compared to the baselines, the results show a 14% increase in AUROC, a 22% decrease in FAR, and an 18% increase in the early detection rate. These findings indicate that there was an improvement in clinical reliability and support for decisions for neurological intensive care unit monitors.References
Jiang, Y., Li, Q., & Zhang, W. (2025). A Real-Time Signal-Based Wavelet Long Short-Term Memory Method for Length-of-Stay Prediction for the Intensive Care Unit: Development and Evaluation Study. JMIR AI, 4, e71247. DOI: 10.2196/71247
Park, D., So, K., Prabhakar, S. K., Kim, C., Lee, J. J., Sohn, J. H., ... & Won, D. O. (2025). Early warning score and feasible complementary approach using artificial intelligence-based bio-signal monitoring system: a review. Biomedical Engineering Letters, 1-18. DOI: 10.1007/s13534-025-00486-4
Venturini, M., Feremans, L., De Corte, W., & Vens, C. Sequential Rule Analysis of ICU Patient Vital Signals and Alarms. In Proceedings of the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases.
Yoo, D., Tae, Y., Cho, K., Je, H., Kim, D., Son, B., ... & Ahn, S. H. (2025). Deep Learning–Based Early Detection of Major Adverse Cerebral Injuries in Cardiothoracic and Vascular Surgery. medRxiv, 2025-01. https://doi.org/10.1101/2025.01.10.25320368
Dhami, A., Onyeukwu, K. A., Sattar, S., Batra, A., Mostafa, Y., Haris, M., ... & Siddique, M. U. (2025). The Prognostic Performance of Artificial Intelligence and Machine Learning Models for Mortality Prediction in Intensive Care Units: A Systematic Review. Cureus, 17(8). DOI: 10.7759/cureus.90465
Ismail, S., Wardah, Z., Wibowo, A., Pranata, S., & Wahyuni, S. (2025). Predictive Analysis of Clinical Status Assessment of Critical Patients Using Electronic Early Warning System Records with Machine Learning. The Malaysian Journal of Nursing (MJN), 17(1), 21-29. https://doi.org/10.31674/mjn.2025.v17i01.003
Khan, M. M., & Shah, N. (2025). AI-driven wearable sensors for postoperative monitoring in surgical patients: A systematic review. Computers in Biology and Medicine, 196, 110783. DOI: 10.1016/j.compbiomed.2025.110783
Su, J., Tie, X., Wei, Y., Zhou, R., Zou, T., Qin, Y., ... & Yin, W. (2025). Critical care ultrasound: development, evolution, current and evolving clinical concepts in critical care medicine. Frontiers in Medicine, 12, 1622604. DOI: 10.3389/fmed.2025.1622604
Choy, X. Y., Wang, L. R., Henderson, T. C., Li, Z. K., Ng, Y. Y., & Fan, X. Neuro-Markovian Approach to Forecasting Patient Vital Signs from Irregular Time-Series. Available at SSRN 5065782. DOI:10.2139/ssrn.5065782
Zi, S., Borde, H. S. D. O., Rocheteau, E., & Lio, P. (2025). Bridging Graph and State-Space Modeling for Intensive Care Unit Length of Stay Prediction. arXiv preprint arXiv:2508.17554. https://doi.org/10.48550/arXiv.2508.17554
Pavithra, D., Parameswaran, T., Choudhry, M., Amrutha, S., Deepa, T., & Kiruthiga, R. (2024). Application of LSTM networks for continuous patient monitoring and anomaly detection in wearable health devices. Indian J. Sci. Technol, 17(37), 3909-3921. DOI: 10.17485/IJST/v17i37.2600
Maleczek, M., Laxar, D., Kapral, L., Kuhrn, M., Abulesz, Y. T., Dibiasi, C., & Kimberger, O. (2024). A comparison of five algorithmic methods and machine learning pattern recognition for artifact detection in electronic records of five different vital signs: A retrospective analysis. Anesthesiology, 141(1), 32-43. DOI: 10.1097/ALN.0000000000004971
Lee, H. Y., Kuo, P. C., Qian, F., Li, C. H., Hu, J. R., Hsu, W. T., ... & Lee, C. C. (2024). Prediction of in-hospital cardiac arrest in the intensive care unit: Machine learning–based multimodal approach. JMIR Medical Informatics, 12(1), e49142. DOI: 10.2196/49142
Yang, M., Peng, Z., van Pul, C., Andriessen, P., Dong, K., Silvertand, D., ... & Long, X. (2024). Continuous prediction and clinical alarm management of late-onset sepsis in preterm infants using vital signs from a patient monitor. Computer Methods and Programs in Biomedicine, 255, 108335. https://doi.org/10.1016/j.cmpb.2024.108335
Verma, A. A., Stukel, T. A., Colacci, M., Bell, S., Ailon, J., Friedrich, J. O., ... & Mamdani, M. (2024). Clinical evaluation of a machine learning–based early warning system for patient deterioration. CMAJ, 196(30), E1027-E1037. DOI: 10.1503/cmaj.240132
Yang, A. C., Ma, W. M., Chiang, D. H., Liao, Y. Z., Lai, H. Y., Lin, S. C., ... & Wang, C. Y. (2025). Early prediction of sepsis using an XGBoost model with single time-point non-invasive vital signs and its correlation with C-reactive protein and procalcitonin: A multi-center study. Intelligence-Based Medicine, 11, 100242. https://doi.org/10.1016/j.ibmed.2025.100242
Malde, A., Prabhu, V. G., Banga, D., Hsieh, M., Renduchintala, C., & Pirrallo, R. (2025). A Machine Learning Approach for Predicting Maternal Health Risks in Lower-Middle-Income Countries Using Sparse Data and Vital Signs. Future Internet, 17(5), 190. https://doi.org/10.3390/fi17050190
Thatha, V. N., Chalichalamala, S., Pamula, U., Krishna, D. P., Chinthakunta, M., Mantena, S. V., ... & Vatambeti, R. (2025). Optimized machine learning mechanism for big data healthcare system to predict disease risk factor. Scientific Reports, 15(1), 14327. https://doi.org/10.1038/s41598-025-98721-6
Deshmukh, M. T., Wankhede, P. R., Chakole, N., Kale, P. D., Jadhav, M. R., Kulkarni, M. B., & Bhaiyya, M. (2025). Towards intelligent food safety: Machine learning approaches for aflatoxin detection and risk prediction. Trends in Food Science & Technology, 105055. https://doi.org/10.1016/j.tifs.2025.105055
Doneda, M., Lanzarone, E., Giberti, C., Vernia, C., Vjerdha, A., Silipo, F., & Giovanardi, P. (2025). An ECG-based machine-learning approach for mortality risk assessment in a large European population. Journal of Electrocardiology, 88, 153850. https://doi.org/10.1016/j.jelectrocard.2024.153850
Nancy, A. A., Ravindran, D., Raj Vincent, P. D., Srinivasan, K., & Gutierrez Reina, D. (2022). Iot-cloud-based smart healthcare monitoring system for heart disease prediction via deep learning. Electronics, 11(15), 2292. https://doi.org/10.3390/electronics11152292
Ramesh, J., Aburukba, R., & Sagahyroon, A. (2021). A remote healthcare monitoring framework for diabetes prediction using machine learning. Healthcare Technology Letters, 8(3), 45-57. doi: 10.1049/htl2.12010
Mudiyanselage, S. E., Nguyen, P. H. D., Rajabi, M. S., & Akhavian, R. (2021). Automated workers’ ergonomic risk assessment in manual material handling using sEMG wearable sensors and machine learning. Electronics, 10(20), 2558. https://doi.org/10.3390/electronics10202558
Alghieth, M. (2025). DeepECG-Net: a hybrid transformer-based deep learning model for real-time ECG anomaly detection. Scientific Reports, 15(1), 20714. https://doi.org/10.1038/s41598-025-07781-1
Wang, L., Yin, Y., Glampson, B., Peach, R., Barahona, M., Delaney, B. C., & Mayer, E. K. (2024). Transformer-based deep learning model for the diagnosis of suspected lung cancer in primary care based on electronic health record data. EBioMedicine, 110. DOI: 10.1016/j.ebiom.2024.105442
Mei, Y., Jin, Z., Ma, W., Ma, Y., Deng, N., Fan, Z., & Wei, S. (2024). Optimizing Acute Coronary Syndrome Patient Treatment: Leveraging Gated Transformer Models for Precise Risk Prediction and Management. Bioengineering, 11(6), 551. DOI: 10.3390/bioengineering11060551
Zisser, M., & Aran, D. (2024). Transformer-based time-to-event prediction for chronic kidney disease deterioration. Journal of the American Medical Informatics Association, 31(4), 980-990. DOI: 10.1093/jamia/ocae025
Zhou, S., Guan, C., Deng, S., Zhu, Y., Yang, W., Zhang, X., ... & Huang, H. (2025). A novel sequence-based transformer model architecture for integrating multi-omics data in preterm birth risk prediction. npj Digital Medicine, 8(1), 536. https://doi.org/10.1038/s41746-025-01942-2
El-Rashidy, N., Sultan, Y. A., & Ali, Z. H. (2025). Predecting power transformer health index and life expectation based on digital twins and multitask LSTM-GRU model. Scientific Reports, 15(1), 1359. https://doi.org/10.1038/s41598-024-83220-x
Moody, B., Hao, S., Gow, B., Pollard, T., Zong, W., & Mark, R. (2022). MIMIC-IV Waveform Database (version 0.1.0). PhysioNet. RRID:SCR_007345. https://doi.org/10.13026/a2mw-f949
DOI:
https://doi.org/10.31449/inf.v50i9.12162Downloads
Published
How to Cite
Issue
Section
License
Authors retain copyright in their work. By submitting to and publishing with Informatica, authors grant the publisher (Slovene Society Informatika) the non-exclusive right to publish, reproduce, and distribute the article and to identify itself as the original publisher.
All articles are published under the Creative Commons Attribution license CC BY 3.0. Under this license, others may share and adapt the work for any purpose, provided appropriate credit is given and changes (if any) are indicated.
Authors may deposit and share the submitted version, accepted manuscript, and published version, provided the original publication in Informatica is properly cited.
