Spatio-temporal graph convolutional networks for daily asthma symptom prediction using passive monitoring of environmental data

Asthma is a chronic disease affecting millions worldwide, with symptoms influenced by personal and environmental factors. To enhance personalized healthcare, we propose a Spatio-Temporal Graph Convolutional Network (ST-GCN) that integrates patients’ static characteristics, environmental factors, and biosignal data to predict daily asthma symptoms. We also introduce a temporal convolutional layer to process high-frequency biosignal data and fuse it with daily environmental inputs. Using data from 22 asthma patients over six months, ST-GCN achieved the highest accuracy of 91.9%, outperforming models like Random Forest, AdaBoost, and Neural Networks. Even without inhaler usage data, ST-GCN maintained an accuracy of 86%, offering a scalable, accessible solution for real-world asthma prediction. Our approach demonstrates the potential of ST-GCN in personalized preventive and prognostic healthcare.