Abstract
Parkinson’s Disease (PD) is one of the most common neurodegenerative conditions, especially among older people. It mainly affects movement and causes symptoms such as tremors, slow walking, muscle stiffness, and balance problems. These symptoms are caused by the gradual loss of brain cells involved in controlling movement. Although PD cannot be cured, treatment and monitoring in the early stages can help people manage their symptoms better and stay independent for longer.Currently, the assessment of PD symptoms is mostly done by clinicians during hospital or clinic visits. These assessments often rely on observation and expert judgement. This makes them subjective and sometimes inconsistent. Also, because such assessments happen occasionally, they may miss important changes in the person’s condition. This can lead to delayed care, lower quality of life, and higher healthcare costs.
This thesis aims to develop an objective and technology-based method to assess PD symptoms by combining brain and body signals. The brain activity is measured using a technique called functional near-infrared spectroscopy (fNIRS), which is non-invasive and can be used while people are moving. At the same time, body motion is recorded using small wearable devices called inertial measurement units (IMUs). These sensors are placed on the body and can measure walking patterns, and movement quality.
The study involves people diagnosed with PD and a group of older adults without PD as the control group. Both groups perform a range of functional walking and mobility tasks that reflect real-life situations. The collected brain and body data are processed and analysed to identify measurable patterns that consistently differ between the two groups. These patterns, known as potential digital biomarkers, may reflect altered brain activation, reduced coordination, or changes in gait patterns specific to PD. To classify these patterns and distinguish individuals with PD from those without the condition, several machine learning models are employed, including Support Vector Machine, Random Forest, Extreme Gradient Boosting, and k-Nearest Neighbour. The goal is to train models that can automatically and accurately detect signs of PD in early stages, based on multimodal data. These models not only support classification but also offer insights into which brain-body features are most important for identifying the disease.
The overall aim of this research is to contribute to the development of an accessible, reliable, and obvjective tool for early-stage diagnosis and ongoing monitoring of PD. By identifying meaningful biomarkers and combining wearable sensors with artificial intelligence, this work has the potential to improve clinical decision-making and enhance the quality of life for people living with PD.
| Date of Award | 2025 |
|---|---|
| Original language | English |
| Supervisor | Maryam GHAHRAMANI (Supervisor), Elisabeth PRESTON (Supervisor) & Raul Fernandez Rojas (Supervisor) |