Biomedical Engineering Reference
In-Depth Information
ratio. In contrast, the present study introduces a quantitative symmetry measure based
on kinematic data which can be expressed as one index.
The Gillette Functional Assessment Questionnaire (GFAQ) Walking Scale is a
widely accepted gait normality measure based on observation. Considerable efforts
have been put into deriving an equivalent measure from kinematic data. Principal com-
ponent analysis (PCA) on 16 discrete gait variables has been used to create a represen-
tation of the data in a different space. The magnitude of the projection of an abnormal
data set onto this space is used as a normality index, known as the Gillette Gait Index
(GGI) [16]. A very similar PCA approach named the Gait Deviation Index (GDI) has
been introduced by [17]. One advantage of PCA approaches is that they transform the
possibly dependent gait variables into a new set of independent variables. The disad-
vantage is that results cannot be traced back to the original gait variables.
A much simpler method, the Gait Profile Score (GPS) and Movement Analysis Pro-
file (MAP), has been suggested [18]. The MAP is created by taking the root mean square
error (RMS) between a reference joint angle curve and the corresponding curve from a
subject. This creates one normality index for each joint angle curve. A unique index, the
GPS, can be derived by concatenating all joint angle curves end to end, and taking the
RMS of this aggregated curve. Although the GDI presents some nice properties such
as normal distributions across GFAQ levels, the GPS is more easily interpreted because
the original variables suffer no transformations and results are given in degrees. It has
been shown that the GPS correlates significantly with clinical judgment [19].
2.3
Instrumented Gait Analysis
Inertial sensors, such as accelerometers and gyroscopes, can complement MOCAP sys-
tems and OGA by providing quantitative and objective gait measurements outside the
gait lab, and for a fraction of the cost.
Most symmetry measures calculated from inertial sensor data take into account only
discrete spatio-temporal variables, e.g. [20], [21]. Although discrete symmetry indices
have been shown useful, a more informative measure of symmetry may be obtained
using the entire continuous sensor data. Few approaches to calculating symmetry us-
ing continuous accelerometer data have been introduced. One example is an unbiased
autocorrelation method using trunk acceleration data [22]. Although this may provide
a good general estimate of gait symmetry, it lacks information about each individual
limb.
More recently, gyroscopes data obtained from shanks and thighs was used to calcu-
late symmetry using a normalized cross correlation approach [23]. This method seg-
ments and normalizes the data to individual strides. As a result, only the shape of the
signal and not its relative temporal characteristics are taken into account. A symbolic
method for estimating gait symmetry using accelerometers [24] or gyroscopes [25] has
been suggested, which takes into account not only the shape but also the temporal char-
acteristics of the signal. This symmetry measure is used in the present paper.
Based on this symmetry measure, the authors proposed a normality measure based
on symbolized inertial sensor data, described in the present paper. No other normality
measures based on inertial sensor data were found in the literature.
Search WWH ::
Custom Search