Real-time mobile health analytics & interventions pipeline to detect acute events in COPD

TY - CHAP

T1 - Real-time mobile health analytics & interventions pipeline to detect acute events in COPD

AU - Sankesara, Heet

AU - Ranjan, Yatharth

AU - Conde, Pauline

AU - Anthobiani, Malik

AU - Rashid, Zulqarnain

AU - Roy Choudhury, Akash

AU - Stewart, Callum

AU - Zhang, Yuezhou

AU - Porter, Joanna

AU - Hurst, John R.

AU - Dobson, Richard

AU - Folarin, Amos

PY - 2024/11/13

Y1 - 2024/11/13

N2 - We developed a real-time mHealth anomaly detection pipeline using wearable data to identify chronic obstructive pulmonary disease (COPD) exacerbation events. A scalable framework implemented on the RADAR-base platform integrates data from questionnaires, smartphone sensors, and wearables for building models and running real-time inference. This case study was designed to check the feasibility of a COPD real-time intervention system, focused on technical and engineering feasibility rather than on the performance of analysis methods. In this paper, we applied an unsupervised Long-Short-Term Memory Autoencoder model trained on physiological data to detect deviations from normal health states. In this six-month study with 20 COPD patients, the system's real-time predictions triggered exacerbation rating scale(ERS) questionnaires. The results showed that the system could detect anomalies up to nine days before patients began medication, with ERS reports indicating normal exacerbation ratings but elevated symptom levels. The results thus proved that the system captured the worsening of symptoms before exacerbation onset, showing the potential of real-time mHealth data in health monitoring, enabling timely intervention and, thus, improved management and prognosis.

AB - We developed a real-time mHealth anomaly detection pipeline using wearable data to identify chronic obstructive pulmonary disease (COPD) exacerbation events. A scalable framework implemented on the RADAR-base platform integrates data from questionnaires, smartphone sensors, and wearables for building models and running real-time inference. This case study was designed to check the feasibility of a COPD real-time intervention system, focused on technical and engineering feasibility rather than on the performance of analysis methods. In this paper, we applied an unsupervised Long-Short-Term Memory Autoencoder model trained on physiological data to detect deviations from normal health states. In this six-month study with 20 COPD patients, the system's real-time predictions triggered exacerbation rating scale(ERS) questionnaires. The results showed that the system could detect anomalies up to nine days before patients began medication, with ERS reports indicating normal exacerbation ratings but elevated symptom levels. The results thus proved that the system captured the worsening of symptoms before exacerbation onset, showing the potential of real-time mHealth data in health monitoring, enabling timely intervention and, thus, improved management and prognosis.

KW - remote-monitoring

KW - mHealth

KW - intervention

KW - real-time

KW - copd

M3 - Conference paper

BT - 21th EAI International Conference on Mobile and Ubiquitous Systems: Computing, Networking and Services

ER -