Using the IPod to quantify dynamic measures of gait independent
of step detection. An innovative way to classify walking
performance of young, healthy older and cognitive impaired
, J.P. van Campen
, L.H.J. Kikkert
, N. Vuillerme
University of Groningen, University Medical Centre Groningen, Center for
Human Movement Sciences, A. Deusinglaan 1, 9700 AD Groningen,
Department of Geriatric Medicine, MC Slotervaart Hospital, Amsterdam,
Univ. Grenoble Alpes, EA AGEIS, La Tronche,
Universitaire de France, Paris, France
Gait analysis focusing on the dynamics of gait (e.g.,
variability, predictability), could advantageously reveal underlying
mechanisms of decreased gait speed and increase the understanding
of the relationship with age-related cognitive decline. Most gait
variables, however, are based on step detection, which could hamper
their appropriate use in clinical practice. Inaccuracies in automatic
step-detection bias these gait variables due to a shuffling gait, whereas
monitoring step-moments is time-consuming. Therefore, we propose
here gait analysis methods using acceleration signals of an IPod that do
not require step-detection.
Data of 3 min. overground walking of young (n = 25; age
26 ± 5.4), healthy old (n = 25; age 65 ± 5.5) and cognitive impaired old
adults (n = 25; age 82 ± 6.3) were recorded with an iPod TouchG4.
Measures that quantify amplitude, regularity [1
and coupling  of 3D trunk accelerations were calculated. Partial
Least Square Discriminant Analysis was used to assess if these
measures accurately classified the age groups. A Receiver Operating
Characteristic curve examined the models sensitivity and specificity.
Four latent factors explained 57% of the variance between
the groups. The young, older and cognitive impaired groups were
classified with a sensitivity of respectively 88%, 80% and 98% and a
specificity of 90%, 86% and 92% showing strong classification power of
Methods that quantify dynamic gait metrics, based on
trunk accelerations using a simple device as the IPod, e.g. independent
of step-detection, can accurately distinguish population groups and
provide insight into how age and cognition affect gait. This enables
automatic qualitative comprehensive analysis of walking performance
in clinical practice.
 Kosse N, Vuillerme N, Hortobágyi T, Lamoth CJC,
Gait & Posture
2016; 46, 112
 Riva F, Toebes MJP, Pijnappels M, Stagni R, van Dieën JH,
2013; 38, 170
 Moe-Nilssen R, Helbostad JL.
J of Biomech.
2004; 37, 121
 Bisi MC, Stagni R.
Gait & Posture
2016; 47, 37
Oral presentations / European Geriatric Medicine 7S1 (2016) S1