Tracking Parkinson’s Disease

Parkinson’s disease is a degenerative, progressive, neurological condition that impairs movement, postural stability, and lowers quality of life. This disease is the second most common age related condition, and as people get older their likelihood of developing Parkinson’s disease rises. Having an efficient way to monitor the progression of this disease and its rehabilitation therefore becomes increasingly important. A way to do this is by tracking balance control. This can be achieved by measuring postural sway using a force platform. Evidence suggests that there is a relation between postural sway and somatosensory inputs, such as sound. This project therefore studies the impact of different metronome frequencies on the postural sway in an entrainment task to answer the question: To what extent does a change in metronome frequency in a balance task impact postural sway? This is done by executing an experiment where 12 healthy participants were asked to perform balance tasks while standing on a force platform. Halfway through each task the frequency of the metronome sound is changed and possible changes in balance as a result of this change were measured. Importantly, this study operates within the research context where using randomized controlled trials is not feasible because it requires too large of a sample size. Drawing causal inferences in this context therefore is particularly challenging. To address this hurdle, we use a model based on the Bayesian nonparametric discontinuity design (BNDD) framework in order to analyze the data collected during the experiments. While BNDD does not guarantee causality on its own, it provides a framework for robustly comparing the evidence in favor of a discontinuous model (indicating an effect) to that of a continuous model (indicating no effect). BNDD is therefore a promising AI technique to offer a nuanced perspective on causal inference and mitigate the overconfidence often seen in these types of quasi-experimental designs. Besides studying the impact of metronome frequencies on postural sway, we also compare the reliability of the data collected using the force platform, to the data collected using a wearable accelerometer to answer the question: To what extent does a wearable accelerometer attached to the sternum provide a reliable patient-friendly alternative to the force platform?