Temporal ensemble resampling random forest for remote patient monitoring in healthcare 4.0

BackgroundRemote patient monitoring (RPM) is an important part of Healthcare 4.0, which allows continuous evaluation of physiological conditions outside of traditional clinical settings. Longitudinal physiological observations from remote patient monitoring may contain clinically relevant abnormalities that are either unusual measurements or changes relative to a patient's recent physiological state. The conventional machine-learning approaches are faced with the problem of reliable anomaly detection due to class imbalance, measurement variability, redundant features and temporal dependencies.ObjectiveA Temporal Ensemble Resampling Random Forest (T-ERRF) framework is proposed for longitudinal physiological anomaly detection in RPM.MethodsThe proposed framework integrates causal temporal feature construction, discrete feature-subset selection by gradient-based crossover optimization (GBOC), class-resampled ensemble learning, and Random Forest classification. Causal temporal features are built only on past observations of the same patient to capture short-term physiological changes and patient-specific deviations without future observations. GBOC is adapted from its original formulation of continuous optimization to a discrete feature-selection problem with the classification performance as the fitness criterion. Class resampling is used during the training phase to counteract the effect of class imbalance and the final model is an ensemble classification Random Forest. The proposed framework is evaluated on a synthetic longitudinal RPM dataset. The dataset includes physiological measurements, patient characteristics, activity information, temporal variables, derived indicators, and anomaly labels, with patient-level separation between training and testing data to prevent information leakage.ResultsT-ERRF achieved 95.30% accuracy, 90.46% precision, 84.95% sensitivity, 97.82% specificity, 87.62% F1-score and 96.12% ROC-AUC on the independent patient-level test set.ConclusionThe results demonstrate the potential of the proposed integrated framework for longitudinal RPM anomaly detection while highlighting the need for validation on real-world wearable and IoT sensor data.

Authors

Institutions

Publication Details

Journal
Technology and Health Care
Published
2026-11-18
DOI
https://doi.org/10.1177/09287329261508547
Primary Topic
Anomaly Detection Techniques and Applications
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Temporal ensemble resampling random forest for remote patient monitoring in healthcare 4.0

Muneer Khan Mohammed, Mustufa Haider Abidi, Usama Umer, Ihab Mohamed Elsayed Koura
3 links
Technology and Health Care
Anomaly Detection Techniques and Applications
article

Temporal ensemble resampling random forest for remote patient monitoring in healthcare 4.0

Muneer Khan Mohammed, Mustufa Haider Abidi, Usama Umer, Ihab Mohamed Elsayed Koura
article en

Abstract

BackgroundRemote patient monitoring (RPM) is an important part of Healthcare 4.0, which allows continuous evaluation of physiological conditions outside of traditional clinical settings. Longitudinal physiological observations from remote patient monitoring may contain clinically relevant abnormalities that are either unusual measurements or changes relative to a patient's recent physiological state. The conventional machine-learning approaches are faced with the problem of reliable anomaly detection due to class imbalance, measurement variability, redundant features and temporal dependencies.ObjectiveA Temporal Ensemble Resampling Random Forest (T-ERRF) framework is proposed for longitudinal physiological anomaly detection in RPM.MethodsThe proposed framework integrates causal temporal feature construction, discrete feature-subset selection by gradient-based crossover optimization (GBOC), class-resampled ensemble learning, and Random Forest classification. Causal temporal features are built only on past observations of the same patient to capture short-term physiological changes and patient-specific deviations without future observations. GBOC is adapted from its original formulation of continuous optimization to a discrete feature-selection problem with the classification performance as the fitness criterion. Class resampling is used during the training phase to counteract the effect of class imbalance and the final model is an ensemble classification Random Forest. The proposed framework is evaluated on a synthetic longitudinal RPM dataset. The dataset includes physiological measurements, patient characteristics, activity information, temporal variables, derived indicators, and anomaly labels, with patient-level separation between training and testing data to prevent information leakage.ResultsT-ERRF achieved 95.30% accuracy, 90.46% precision, 84.95% sensitivity, 97.82% specificity, 87.62% F1-score and 96.12% ROC-AUC on the independent patient-level test set.ConclusionThe results demonstrate the potential of the proposed integrated framework for longitudinal RPM anomaly detection while highlighting the need for validation on real-world wearable and IoT sensor data.

Technology and Health Care
King Saud University (SA)
Deanship of Scientific Research, King Saud University
Good health and well-being
Anomaly Detection Techniques and Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon