A novel approach for analysis of altered gait variability in amyotrophic lateral sclerosis.

Gait variability reflects important information for the maintenance of human beings' health. For pathological populations, changes in gait variability signal the presence of abnormal motor control strategies. Quantitative analysis of the altered gait variability in patients with amyotrophic lateral sclerosis (ALS) will be helpful for either diagnosing or monitoring pathological progression of the disease. Thus, we applied Teager energy operator, an energy measure that can highlight the deviations from moment to moment of a time series, to produce an instantaneous energy time series. Then, two important features were extracted to assess the variability of the new time series. First, the standard deviation statistics were used to measure the magnitude of the variability. Second, to quantify the temporal structural characteristics of the variability, the permutation entropy was applied as a tool from the nonlinear dynamics. In the classification experiments, the two proposed features were input to the support vector machine classifier, and the dataset consists of 12 ALS patients and 16 healthy control subjects. The experimental results showed that an area of 0.9643 under the receiver operating characteristic curve was achieved, and the classification accuracy evaluated by leave-one-out cross-validation method could reach 92.86 %.