Table 1.
Mean value and standard deviation about the age, height, weight, BMI, sex, and pain (as assessed by the KOOS score) and the number of subjects that have experienced a surgery or an injury for both the control and the knee OA subjects.
Figure 1.
Figure 1a is the real, lab-based environment, Figure 1c is the computer reconstruction, whereas Figure 1b an overlay of the two.
Figure 2.
The blue curve corresponds to the mean GRF curve, whereas the blue shaded region indicates the precision of plus minus one standard deviation.
Accordingly, the foot which has knee OA is depicted in red.
Figure 3.
GRFs for an random indicative subject.
Figure 4.
The goodness of fit of a Gaussian distribution to the actual empirical distribution of the GRFs patterns for (a) normal subjects and (b) knee OA subjects.
Probability distributions over GRF patterns. Solid red lines is Gaussian distributions with mean and standard deviation matched to the empirical GRFs histograms. The data and the matching Gaussian distributions appear as bell-shaped.
Figure 5.
The proposed complexity measure for the first 36 PCs.
In the lower PC dimensional space knee OA subjects have a tendency to present lower values.
Figure 6.
How much variability is explained as a function of the number of the components.
The x-axis corresponds to the number of PCs, whereas the y-axis the percentage of the variance of the GRF patterns explained by the respective number of PCs. It is evident that human walking is a complex process, since the slope starts at a low point (1 PC explains just above 30% of the variability of the combined data) and the slope progresses slowly.
Figure 7.
Projection of the GRF patterns in 2-D PC space (i.e. the first two PCs).
The two classes are not separable.
Figure 8.
PC visualisations as discriminants of the two classes: normal subjects vs. knee OA subjects.