Citation

Trisna, Nusrat Jahan and Mridha, M.F. and Hossen, Md. Jakir (2026) A hybrid clinical temporal graph neural network model for time series data to predict health deterioration in ICU patients. Information Systems, 140. p. 102727. ISSN 0306-4379

Text 2.pdf - Published Version Restricted to Repository staff only Download (4MB)

Abstract

Models based on machine learning (ML) that can predict a patient’s health getting worse are very useful in intensive care units (ICUs) because they can let doctors know when a patient’s condition is getting worse very quickly. But traditional models do not always work well with time-series medical data and do not respond quickly to important changes. This study suggests a Clinical Temporal Graph Neural Network (Clin-TGNN) framework to predict worsening conditions like sepsis and cardiac arrest by using graph structures to model how things change over time. The model is tested on several datasets, including a combined dataset (MIMIC-III, ReXErr-v1, and Two-Tiered) and a standalone MIMIC-IV dataset. The dataset ‘‘Nonlinear Relationship Between Early Postoperative Pulse Pressure Variation and Acute Kidney Injury Risk in Elderly Cardiac Surgical Patients’’ is used to look at discharge timing and length of stay (LOS). The results show that Clin-TGNN works very well. The combined dataset(MIMIC-III + ReXErr-v1 + Two-Tiered datasets) has an accuracy of 0.9600 and an AUC of 0.9980, while the MIMIC-IV dataset has an accuracy of 0.7558 and an AUC of 0.8180. The model has a 0.9687 accuracy rate and a 0.9883 AUC rate, with a 4 h early warning lead time and very few false alarms. This method is very helpful for doctors because it lets them find risks early and plan better for patient care and hospital resources.Clin-TGNN is a strong and flexible framework for keeping an eye on patients in the ICU and predicting their outcomes that works well in the real world.

Item Type:	Article
Uncontrolled Keywords:	ICU monitoring
Subjects:	R Medicine > R Medicine (General) > R855-855.5 Medical technology
Divisions:	Faculty of Engineering and Technology (FET)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	04 May 2026 01:02
Last Modified:	07 May 2026 04:45
URII:	http://shdl.mmu.edu.my/id/eprint/15802

Downloads

Edit (login required)