Monitoring behavioural changes in households with dementia using intelligent ambient sensing technologies
File(s)
Author(s)
Serban, Alina-Irina
Type
Thesis
Abstract
Dementia is characterised by diseases and conditions caused by a decline in cognitive skills affecting a person’s ability to perform activities of daily living. Intelligent ambient sensing technologies are becoming a new standard for long-term, passive monitoring of health-related changes, improving the quality of life for people living with dementia. Current literature explores short-term patterns and lab-based studies, yet these lack the real-time, long-term aspects of living with dementia.
The work presented in this thesis explores data engineering and machine learning methods to develop novel applications for detecting behavioural changes in households with dementia. The data is collected from unobtrusive sensors which detect motion in the house, door usage, appliance usage and bed occupancy. This set-up exists in over 100 households with dementia that are part of the UK Dementia Research Institute Care Research and Technology study since 2019.
This thesis comprises of four studies. Design engineering techniques were used to identify the requirements for developing responsible and applicable at-home monitoring solutions in a user-centred approach. The COVID-19 pandemic created a naturalistic experimental set-up during which changes in activity around the house and going-out habits were monitored. A novel algorithm was developed to recognise patterns of behaviour associated with going outside at night to monitor night-time safety (algorithm under validation as part of the Minder study). The last analysis investigated the relation between activity in the house, layout and demographic details to extract unique household fingerprint behaviours that correctly identify individual households at the group level.
The studies presented in this thesis demonstrate the use of intelligent ambient sensing technologies to monitor real-life changes in behaviour of households living with dementia. These changes can be used to identify markers related to disease progression or clinical events, which can help guide clinical assessment and intervention.
The work presented in this thesis explores data engineering and machine learning methods to develop novel applications for detecting behavioural changes in households with dementia. The data is collected from unobtrusive sensors which detect motion in the house, door usage, appliance usage and bed occupancy. This set-up exists in over 100 households with dementia that are part of the UK Dementia Research Institute Care Research and Technology study since 2019.
This thesis comprises of four studies. Design engineering techniques were used to identify the requirements for developing responsible and applicable at-home monitoring solutions in a user-centred approach. The COVID-19 pandemic created a naturalistic experimental set-up during which changes in activity around the house and going-out habits were monitored. A novel algorithm was developed to recognise patterns of behaviour associated with going outside at night to monitor night-time safety (algorithm under validation as part of the Minder study). The last analysis investigated the relation between activity in the house, layout and demographic details to extract unique household fingerprint behaviours that correctly identify individual households at the group level.
The studies presented in this thesis demonstrate the use of intelligent ambient sensing technologies to monitor real-life changes in behaviour of households living with dementia. These changes can be used to identify markers related to disease progression or clinical events, which can help guide clinical assessment and intervention.
Version
Open Access
Date Issued
2024-01-08
Date Awarded
01/06/2024
Copyright Statement
Creative Commons Attribution NonCommercial Licence
License URL
Advisor
Calvo, Rafael
Sharp, David
Sponsor
UK Research and Innovation
Grant Number
Grant No. P/S023283/1
Publisher Department
Computing
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)