Multi-scale graph modeling and analysis of locomotion dynamics towards sensor-based dementia assessment

Abstract

Dementia is a general neurodegenerative disorder beyond normal aging, which is not only overwhelming for the patients, but also negatively affects their caregivers and families. In the state of the art, paper-based survey methods such as the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) are widely used for the assessment of dementia conditions. However, these methods require lab visits or administration from nurses, physicians and examiners, and are limited in the ability to track temporal degradation (or daily variations) of dementia conditions. With rapid advances in sensing technology, there is growing interest in the development of new, sensor-based methods that provide more flexibility in dementia monitoring and require minimal interventions from practitioners. In this article, we propose a new, sensor-based method that estimates dementia conditions with daily locomotion data. The proposed methodology is evaluated and validated with both simulation and real-world case studies. Experimental results show the compelling predictive accuracy, with both true positive and true negative rates above 85%. This article shows that sensor-based methods have great potential for real-time monitoring of temporal variations of dementia conditions from daily gait locomotion dynamics.

Keywords:

• Dementia
• graph model
• locomotion dynamics
• Mini-Mental State Examination
• multi-scale analysis
• paper-based survey
• wireless sensing

Acknowledgments

The authors would like to thank Dr. William Diehl Kearns and James Leonard Fozard School of Aging Studies, University of South Florida, Tampa, FL, USA, for sharing the data for this investigation.

Additional information

Funding

This work is supported by the National Science Foundation (CMMI-1646660). The authors also thank the Harold and Inge Marcus Career Professorship (HY) for additional financial support.