Assessment of Gait Symmetry and Gait Normality Using Inertial Sensors: In-Lab and In-Situ Evaluation - Biomedical Engineering Systems and Technologies - page 222

to calculate the Spearman's rank correlation coefficient with kinematic measurements,
MAP, GPS, MAP-symmetry and GPS-symmetry. The optimal number of symbols was
chosen so as to maximize the correlation coefficients. These optimal parameters were
used to calculate symmetry and normality for the hip-replacement patients.
The Spearman's rank correlation coefficient was used to evaluate the correlation be-
tween two variables. The non-parametric Wilcoxon rank sum test was used to compare
two distributions, and a Kruskal-Wallis test was used to compare more than two distri-
butions. All linear model approximations were calculated based on least mean square
errors.
The area under the receiver operating characteristic curve (AUC) was used to eval-
uate the discriminatory power of the normality index. The ROC curve was constructed
based on tests performed on the same individuals. Therefore, any statistically signifi-
cant comparison between different AUC must take into account the correlated nature
of the data. A nonparametric approach based on generalized U-statistics was used to
estimate the covariance matrix of the different curves [27].
All measurements of the in-situ data collection included two trials, which were used
to assess the test-retest reliability of each index using intra-class correlation coefficient
(ICC) type A-1 as a measure of absolute agreement [28]. All tests were bi-directional
with confidence level, α =0 . 05 . All data analysis was undertaken in MATLAB (Math-
Works, Natick, MA).

4R su s

4.1 In-Lab Evaluation

The best correlation of MAP and GPS with the proposed normality index was achieved
with the waist accelerometer sensor, 18 symbols and transition histograms. The best cor-
related signals are shown in Table 1. The best correlation of the inertial sensor symmetry
with MAP-symmetry and GPS-symmetry was achieved with the shank gyroscopes, 20
symbols, and symbol period histograms. The best correlation coefficients are shown in
Table 2. These configurations were used to calculate normality and symmetry respec-
tively in the in-situ evaluation.

4.2 In-Situ Evaluation

All but one participant answered the EQ-5D TM questionnaire on both occasions. The
values of the answers given to each category were added to a single score for that
category. Results from before the operation and after the follow-up session are shown
in Figure 4. Lower scores correspond to more patients in better health. The biggest
changes were regarding mobility, usual activities and pain/discomfort.
Symmetry results for baseline and follow-up measurements are shown in Figure 5
for each subject. Measurements were averaged over both trials of each session. The
symmetry index ranges from 0 to 100, a low symmetry index indicates good symme-
try whereas a high value indicates asymmetry. According to the proposed index, gait
symmetry improved at follow-up for approximately half the subjects. The asymmetry
at follow-up may be caused by the use of one crutch. The symmetry index according