A semantic digital health pipeline for interpretable and longitudinal gait monitoring using wearable sensors
4 affiliations
- 1Departamento de Ingeniería de Organización, Administración de Empresas y Estadística, Escuela Técnica Superior de Ingenieros Industriales, Universidad Politécnica de Madrid (UPM), C/ José Gutiérrez Abascal, Madrid, Spain
- 2Facultad de Medicina, Salud y Deporte, Universidad Europea de Madrid, Villaviciosa de Odón, Madrid, Spain
- 3Unidad de Esclerosis Múltiple, Hospital Universitario de Getafe, Getafe, Spain
- 4Instituto de Investigación Sanitaria del Hospital Universitario La Paz (IdiPAZ), Madrid, Spain.
Abstract
Background: Wearable gait monitoring can support neurological follow-up, but raw inertial time-series are difficult to interpret clinically.
Objective: To evaluate a semantic pipeline that converts raw wearable streams into structured gait events for free-living monitoring.
Methods: Instrumented smart socks captured tri-axial accelerometer and gyroscope signals from a bilateral inertial measurement unit (IMU) configuration. Data were stored as raw time-series. A deterministic rule-based pipeline derived time-bounded events (per-limb sensor-reception intervals, concurrent bilateral candidate intervals, per-limb effective-movement intervals, and bilateral gait episodes) in a relational database. The semantic layer was published as a Web Ontology Language (OWL) ontology with an Ontop ontology-based data access (OBDA) mapping and reusable SPARQL Protocol and RDF Query Language (SPARQL) queries. Feasibility was evaluated in 44 participants (31 people with multiple sclerosis (PwMS) and 13 controls) across 257 monitoring sessions.
Results: The pipeline identified 6,224 gait episodes (∼60.8 hours walking). Episode counts were right-skewed (mean 141.5 episodes/participant, 95% CI 109–174). Most participants (80%) showed low left–right asymmetry (mean index 0.069, 95% CI 0.040–0.107), while elevated imbalance was concentrated in low-data participants. Against standardised in-clinic tests, positive-interval detection sensitivity was 85% (50/59 intervals with bilateral data) after per-session clock alignment using the same annotations; this potentially optimistic estimate does not establish specificity or free-living false-positive performance. In the frozen analysis snapshot, semantic-consistency checking confirmed that 100% of the 6,224 reported gait episodes were temporally contained within concurrent bilateral candidate intervals (activity_all).
Conclusions: Hierarchical semantic temporal events enable transparent, scalable, and interpretable real-world gait monitoring. By releasing the ontology, OBDA mapping, and queries, the pipeline supports longitudinal assessment, limb balance analysis, and interoperable health-system integration. The same approach is applicable beyond gait to other wearable-derived digital biomarkers and rehabilitation-monitoring tasks. This work does not deliver a validated clinical gait assessment tool, but demonstrates reusable digital health representations.
Related
- An interpretable ANFIS-NSGA-II framework for evidence-stratified modular wearable risk fusion in elder-health monitoring
- From wetting-front coalescence to urban road collapse: interpretable machine learning of hydro-mechanical failure under coupled rainfall and pipeline leakage
- Wearable Estimation of Knee Joint Contact Forces: SiMPLeSENSE, a Simulated-IMU Pipeline Validated Against In Vivo Loads
- Trustworthy Stress Detection from Wearable Sensor Signals for Preventive Anxiety Monitoring: Subject-Independent Accuracy and Calibrated Confidence with the Autonomic-Branch Quantum-Graph Network
- IOT-enabled wearable device and cloud framework for chronic disease identification and health monitoring using LST-BCLNM
BibTeX
@article{farmanesh2026semantic,
title = {{A semantic digital health pipeline for interpretable and longitudinal gait monitoring using wearable sensors}},
author = {Amir Farmanesh and Joaquín Ordieres-Meré and Raúl G. Sanchis and Yolanda Aladro-Benito},
journal = {Digital Health},
year = {2026},
month = sep,
doi = {10.1177/20552076261488885},
url = {https://doi.org/10.1177/20552076261488885},
note = {CC BY-NC 4.0},
}
RIS
TY - JOUR
TI - A semantic digital health pipeline for interpretable and longitudinal gait monitoring using wearable sensors
AU - Amir Farmanesh
AU - Joaquín Ordieres-Meré
AU - Raúl G. Sanchis
AU - Yolanda Aladro-Benito
PY - 2026
DA - 2026/09/16
JO - Digital Health
DO - 10.1177/20552076261488885
UR - https://doi.org/10.1177/20552076261488885
AB - Background: Wearable gait monitoring can support neurological follow-up, but raw inertial time-series are difficult to interpret clinically. Objective: To evaluate a semantic pipeline that converts raw wearable streams into structured gait events for free-living monitoring. Methods: Instrumented smart socks captured tri-axial accelerometer and gyroscope signals from a bilateral inertial measurement unit (IMU) configuration. Data were stored as raw time-series. A deterministic rule-based pipeline derived time-bounded events (per-limb sensor-reception intervals, concurrent bilateral candidate intervals, per-limb effective-movement intervals, and bilateral gait episodes) in a relational database. The semantic layer was published as a Web Ontology Language (OWL) ontology with an Ontop ontology-based data access (OBDA) mapping and reusable SPARQL Protocol and RDF Query Language (SPARQL) queries. Feasibility was evaluated in 44 participants (31 people with multiple sclerosis (PwMS) and 13 controls) across 257 monitoring sessions. Results: The pipeline identified 6,224 gait episodes (∼60.8 hours walking). Episode counts were right-skewed (mean 141.5 episodes/participant, 95% CI 109–174). Most participants (80%) showed low left–right asymmetry (mean index 0.069, 95% CI 0.040–0.107), while elevated imbalance was concentrated in low-data participants. Against standardised in-clinic tests, positive-interval detection sensitivity was 85% (50/59 intervals with bilateral data) after per-session clock alignment using the same annotations; this potentially optimistic estimate does not establish specificity or free-living false-positive performance. In the frozen analysis snapshot, semantic-consistency checking confirmed that 100% of the 6,224 reported gait episodes were temporally contained within concurrent bilateral candidate intervals (activity_all). Conclusions: Hierarchical semantic temporal events enable transparent, scalable, and interpretable real-world gait monitoring. By releasing the ontology, OBDA mapping, and queries, the pipeline supports longitudinal assessment, limb balance analysis, and interoperable health-system integration. The same approach is applicable beyond gait to other wearable-derived digital biomarkers and rehabilitation-monitoring tasks. This work does not deliver a validated clinical gait assessment tool, but demonstrates reusable digital health representations.
SN - 2055-2076
ER -