Sensor-Based Human Motion Analysis and Machine Learning for Dementia Detection

Dementia is a progressive neurodegenerative condition that affects memory, cognition, behaviour and the ability to perform everyday activities. It currently affects more than 55 million people worldwide, with this number projected to rise to approximately 150 million by 2050. Delayed or inaccurate diagnosis remains common, reducing opportunities for early intervention and contributing to growing personal, social and healthcare burdens. Emerging evidence suggests that subtle changes in movement, gait, posture and balance may occur during the early stages of dementia, sometimes before significant memory decline becomes apparent. These changes could provide objective digital indicators of cognitive and functional deterioration.

This PhD project aims to develop a non-invasive, sensor-based approach for dementia detection using human motion and balance data. Movement-related features will be extracted from data collected during carefully designed motor and cognitive-motor tasks. Statistical analysis and machine-learning methods will then be used to identify digital biomarkers associated with dementia and distinguish people living with dementia from cognitively healthy individuals. The project will also investigate which sensing technologies, assessment tasks and movement features provide the most reliable and clinically meaningful information. The anticipated outcome is an objective and accessible assessment framework that could complement existing clinical evaluations, support earlier identification of dementia-related impairment and inform timely intervention and care.