|rgu.ac.uk||AtoZ | Contact | Search | Intranet | Moodle | Student Portal|
|Home | Support | Research | Staff | Contact us|
|RGU > School of Computing|
Visualising Human Gait DataDr Robert Noble and Dr Ray White, Sunderland University
Human "gait" data refers to the pattern of movements generated by various joints in the human body when an action is performed. Our research has focussed on data captured from the leg joints of subjects when walking. In particular, graphs may be plotted of the joints in the leg during the walking cycle. Adults who walks "normally" and have no leg problems will generate similar graphs. By measuring many such adults a mean graph can be found for each joint. Similarly, a range, based on standard deviations from the mean, can be found. Graphs for a "normal" subject would then be expected to lie within the mean plus or minus two standard deviation curves.
The mean value is shown dotted with the standard deviation limits as solid grey curves. The left and right legs are shown in red and green. Along the top is a thermometer scale after Manal and Stanhope (2004), showing the deviation from the normal curve, in this case for the left leg.
A subject is measured using special equipment. In our case this has been using the Oxford Metric Vicon system. Here, markers are placed on the subject and tracked by cameras. The computer can then calculate the joint movements from the markers. Force plates in the floor allow forces and hence moments and power to be found as well.
For a given subject, graphs can be generated for the ankle, knee, hip and pelvis, for both right and left leg. Graphs can be of side, front or top view and can be plotted as angle, force, moment or power. This gives a total of 96 possible graphs.
For subjects who have walking problems, it would be expected that parts of some of the graphs will be outside the expected range. A skilled clinician requires a considerable amount of time to analyse this data. In addition to this, simply sifting through 96 possible graphs is time consuming.
The aim of this research is to provide the clinician with a tool that will direct the clinician to the meaningful parts of the data and thus speed up the analysis. The approach used was to utilise the methods of Information Visualisation (see for example, Spence 2001, Card et. al. 1999).
When data from a subject was read in by the system, a display is generated showing potentially interesting areas, that is, parts of graphs which are out of range, as shown here:
This display will be for a particular leg (left), a particular view (side) and a particular type (angle). These can be changed by the clinician by clicking an appropriate button.
The rows represent joints and the columns represent specific events or patterns that are meaningful to the clinician. First, the events which represent either foot hitting or leaving the floor. Then the position and size of various peaks in the graph.
The arrows in the boxes give a visual representation of how out of range the given curve is at the given event. This gives an immediate oversight into the data, allowing a particularly interesting feature to be spotted. For example, it is clear in the above that the pelvic (top) graph is too high throughout. The corresponding graph can be displayed simply by clicking on the icon of the pelvis at the left hand side.
Experience has shown that certain graphical features may well relate to specific problems. Although no universally agreed set has been defined, our research incorporates this idea, taking some suggestions from Gibbs S (2000). We do not claim these are always applicable, but we do claim that our technique demonstrates how any such rule which a clinician finds useful can be incorporated into the program to speed up diagnosis. The burden of diagnosis lies with the clinician, but the use of the rules may help speed this up.
The basic idea is that the rules pick out features of the graphs which may indicate a particular problem. When a subject's graphs are loaded, any rule which fires results in the corresponding graph or graphs being displayed, along with indications as to why they fired. Text below the graph also states why the graph is there. Shown here is a rule for which the graph is too high (hence the up arrow) just before toe off (the part of the graph highlighted in red).
A given problem may be associated with several rules and may only be considered if all of them fire. For this reason there is a "hypothesis" button which allows the set of rules corresponding to a given problem to be run. All corresponding graphs are shown, with any deviations that match the rules. If one or more graphs has no matches it is shown without highlights or arrows. From this, the clinician can draw suitable conclusions.
The complete interface is as shown below:
At this point the prototype is running, as shown here. The next step is to gather suitable data and to start to put together some more widely accepted rules.
Further details can be found in Noble & White 2005 and Noble & White 2008.
Card SK, Mackinlay JD, Shneiderman B (Eds). Readings in Information Visualization: Using vision to think. San Francisco, Morgan Kaufmann 1999.
Gibbs S (2000) Guide to the Interpretation of Gait Deviations, The Dundee Royal Infirmary, Scotland, Gait Course 2000, 4th — 7th April, 2000.
Manal K, Stanhope SJ. A novel method for displaying gait and clinical movement analysis data. Gait Posture 2004; 20: 222-226.
Noble R, White R. Visualisation of Gait Analysis Data. 9th International Conference, Information Visualisation, London, 6-8th July 2005 pp 247-252
Noble R, White R. Reporting Clinical Gait Analysis data, in User Centered Design for Medical Visualization edited by Feng Dong, Gheorghita Ghinea, and Sherry Y. Chen, ISBN-10: 1599047772, ISBN-13: 978-1599047775
Spence R. Information Visualization, Essex: ACM Press, 2001.
|Disclaimer | Freedom of Information | Code of Conduct | ©||School of Computing, The Robert Gordon University|