Unveiling Interpretable Key Performance Indicators in Hypoxic Response: A System Identification Approach

Jing Chen,Rong Xiao,Lei Wang,Lingling Zhu,Dawei Shi

DOI: https://doi.org/10.1109/tie.2021.3137618

IF: 7.7

2021-01-01

IEEE Transactions on Industrial Electronics

Abstract:Advances in wireless networks and wearable electronic devices have enabled the monitoring of massive physiological data for disease diagnosis. In this article, we aim to explore the hypoxic response dynamics through the use of system identification based on physiological data monitored by wearable devices. A third-order autoregressive moving average with exogenous inputs model is developed to describe the dominant system dynamics, based on which an interpretable index called sum of distance (SoD) is proposed from a systems and control perspective for AMS risk evaluation. The effectiveness of SoD is evaluated on the basis of physiological data from a proof-of-the-concept study. Statistically, significant relationships of DSI with ground truth AMS metrics (including, Lake Louise score, deep sleep duration, and deep sleep ratio) are observed. To accelerate the evaluation algorithm design, a simulator is designed. A model parameter selection method based on the three-sigma rule is proposed to generate an in silico population, and a total disturbance sequence is determined for disturbance simulation. The created data has the same trend as the real measurements. The proposed method and experimental results indicate the feasibility of improving the AMS risk evaluation performance by understanding and learning the hypoxic response mechanism.