Postural Sway and Machine Learning

Research areas


Postural sway is the measurement of a person's (or animal's) centre of pressure (CoP) as they stand still on a force plate, with eyes open or closed (see Figure 1). Sway can also be measured using inertial sensors or with video cameras. Because this is a relatively straightforward output from a complex, non-linear system (postural control), it is of interest in several areas where this system could be disrupted, such as Parkinson's Disease, schizophrenia, injury/concussion, and normal ageing. However, there are multiple ways of quantifying postural sway, ranging from the simple (sway path length, sway area) to the more sophisticated (recurrence quantification analysis, detrended fluctuation analysis - based on principles of nonlinear dynamics), and these are used in an ad hoc fashion in different studies. 

Figure 1: top) An illustration of postural sway in the mediolateral (ML) and anteroposterior (AP) directions; middle) A graphical depiction of how the centre of pressure (CoP) is derived; bottom) some real data from a participant with eyes open and eyes closed.


Using a combination of supervised and unsupervised machine learning techniques, with a fixed set of postural sway measures, we are interested to determine which features are the data are most informative in discriminating between different datasets: 

  • Eyes open and eyes closed conditions in the same individuals
  • Older and younger adults
  • PD patients and age-matched controls
  • Individuals with schizophrenia or schizotypal personality disorder (SPD) and age-matched controls
  • Long-term cannabis users and age-matched controls
We already have data for most of these questions, so this project would be purely data-based, although it has the potential to produce results which call for experimental replication in a new sample. 


This project will appeal to students with excellent skills in experimentation, programming, and teamwork. The preference is for students who have finished/are taking the units of Artificial Intelligence and/or Machine Learning.

Background Literature

Apthorp, D., Nagle, F., & Palmisano, S. (2014). Chaos in balance: non-linear measures of postural control predict individual variations in visual illusions of motion. PLoS ONE, 9(12), e113897.

Ramdani, S., Tallon, G., Bernard, P. L., & Blain, H. (2013). Recurrence Quantification Analysis of human postural fluctuations in older fallers and non-fallers. Annals of Biomedical Engineering, 41(8), 1713-25.
Kent, J. S., Hong, S. L., Bolbecker, A. R., Klaunig, M. J., Forsyth, J. K., O'Donnell, B. F., & Hetrick, W. P. (2012). Motor deficits in schizophrenia quantified by nonlinear analysis of postural sway. PLoS ONE, 7(8), e41808.


The student will develop an understanding of the applications of machine learning to human physiological and medical data, which is a rapidly growing field with many opportunities for new developments and discoveries. There is also potential for inclusion on published papers and conference travel to present them if the student's contribution is sufficient. Further involvement in our ongoing project on Parkinson's disease in collaboration with Stanford University and TCH is also a possibility.

Updated:  1 June 2019/Responsible Officer:  Dean, CECS/Page Contact:  CECS Marketing