Utilizing graph neural networks for adverse health detection and personalized decision making in sensor-based remote monitoring for dementia care

Nivedita Bijlani, Oscar Mendez Maldonado, Ramin Nilforooshan, Payam M. Barnaghi, Samaneh Kouchaki

Abstract: Highlights•We propose a GNN approach to detect unusual activity in the homes of people living with dementia.•We compute personalized alert thresholds guided by clinician-defined target alert rates.•We introduce the first use of negative sample-free graph contrastive learning in healthcare.•Our approach outperforms SOTA temporal graph models in accuracy and speed across 3 patient cohorts.