Short communication
Statistical analysis of the 180 degree walking turn: Common patterns, repeatability and prediction bands of turn signals

https://doi.org/10.1016/j.bspc.2019.101689Get rights and content

Highlights

  • Angular rate along roll, pitch, and yaw axes showed different level of reliability.

  • Translational acceleration exhibited poor intra-subject repeatability.

  • Gaussian point-by-point method showed lower true coverage than bootstrap method.

Abstract

Turning is an essential movement and has been shown to be a relevant measure for differentiating pathologies. Nowadays, turn analyses utilizing inertial measurement units (IMU) have expanded. Although several IMU-based turn metrics exist, there is no information on the repeatability of turn signals and on the existence of signal patterns shared across subjects. Also, the variability of IMU signals within various subject groups has not been estimated yet. This paper presents an analysis of turn angular velocity and acceleration provided by IMU and tests them for repeatability, patterns, and variability within groups of healthy and diseased subjects. Intra-class correlation and methods for estimating prediction bands, namely the Gaussian point-by-point and bootstrap method, were employed to analyze turn signals from Parkinson disease patients and a control group. The yaw angular velocity demonstrated the highest repeatability in both groups as well as reliability of a shared pattern (p = 0.79 and 0.86). The bootstrap method showed wider bands and higher true coverage in comparison to its Gaussian counterpart. From the results of the performed analysis, we recommend the bootstrap method for determining prediction bands. We also recommend the yaw angular velocity as the signal to be assessed in turn analysis.

Introduction

Turning is an essential movement which is required in nearly every daily activity task. Turning is a challenging task and there is evidence that turns are associated with a higher risk of falling [1]. Previous studies indicate that quantitative parameters of turning are useful markers that are sensitive to age effects and different pathologies. Namely, they were shown to differentiate healthy elderly from those with a mild cognitive impairment [2] or PD patients [3], and elderly adults with daily living disabilities [4]. A need to consider turning manoeuvres in routine clinical practice has been suggested [5].

Widely used data acquisition devices output continuous curves. These curves are expressed as a function of the time or percentage of turn. Usually, a single parameter is computed from the continuous curve. Extracted parameters are typically a minimal, maximal, mean value, or the value at a specific event (e.g. at heel strike). In comparison to single parameter analysis, analysis of the continuous curve is more informative [6].

In research and clinical practice, four types of issues related to curve analysis are encountered. The first issue concerns whether the curves of subject groups have the same pattern. This is important in selecting parameters for quantification of the turning manoeuvre because the quantification of curves with different patterns may not be adequate. Moreover, when the curves have a similar pattern then they can be processed to build the mean curve which serves as a representative curve for a given subject group.

The second issue concerns the repeatability of curves. When a turn task is performed repeatedly, e.g. within one session or an inter-session, the recorded curves should be characteristic for the subject to represent a usable parameter.

The third problem deals with classification of individual curves, i.e. deciding whether the subject's curve belongs to specific population (e.g. subject groups) or not. Statistical methods used in analysis of single parameters are not suitable for continuous curves. It has been shown that prediction bands are an adequate statistical tool when applied to continuous curves of gait [[7], [8], [9], [10]], cervical spine movement [11], and scapulo-humeral coordination [12]. Using prediction bands, the range of likelihood kinematics of subject groups can be defined. Determination of a “normal” range is necessary for researchers and clinicians to classify whether assessed curves belong to the same population as the training curves.

Finally, the fourth problem covers the comparison of subject groups. The comparison of the mean differences between subject groups can be useful to designate attractive parts which differentiate between the groups. Such differences might lay a foundation for further quantitative analysis.

An increasing amount of research involving the 180° walking turn [[13], [14], [15], [16], [17]] utilizes inertial measurement units to acquire data about the translational and/or rotational component of movement. However, the suitability, i.e. common pattern, of various turn signals for continuous analysis have not been explored yet. Furthermore, the intra-subject repeatability of curves has not been determined. Regarding the confidence bands of the walking turn manoeuvre, there was only one study that presented confidence bands for lower limbs kinematics [18] and kinetics [19] in typically developing children. It would be beneficial to researchers if a range of normal kinematics, namely angular velocity and acceleration, were established for the 180° turn manoeuvre.

Based on the highlighted problems, we address the following aims: (1) to identify which signals are suitable for walking turn analysis (2) to calculate prediction bands for walking turn signals, (3) provide reference curves for two subject groups, namely older adults and Parkinson disease patients, and (4) suggest promising turn segments for further assessment.

Section snippets

Participants and data acquisition

In the study we included 27 older healthy volunteers (24 males, 3 females), mean age 64.2 (SD 8.3) years without history of neurological disorder, and 24 mild treatment-naive Parkinson disease (PD) patients (15 males, 9 females), mean age 59.2 (SD 11.9) years. The study was approved by the Ethics Committee of the General University Hospital in Prague, Czech Republic, and therefore performed in accordance with the ethical standards established in the 1964 Declaration of Helsinki. Written,

Common curve patterns

The inter-subject reliability of all acceleration curves was poor in both subject groups (ρ < 0.50). For angular velocity signals the reliability varies from poor (ρ < 0.50) to good (0.90 > ρ > 0.75). Yaw angular velocity of both groups demonstrated good reliability (ρ = 0.79 for PD, ρ = 0.86 for CG). For details refer to Table 1.

Intra-subject repeatability

The intra-subject reliability of the yaw angular velocity varies from moderate (ρ > 0.50) to excellent (ρ > 0.90). Other angular velocity curves showed reliability

Discussion

The current study employed statistical methods to analyse walking turn curves from an instrumented Timed-Up and Go test. Using intra-class correlation this study provides assessment of within-group common patterns and intra-subject repeatability of turn angular velocity and acceleration. Next, the range of the likelihood of the turn kinematics of the subject group was determined via prediction bands. Additionally, this study provides a unique comparison of turning curves from different subject

Acknowledgements

This work was supported by Czech health research council (Ministry of Health of the Czech Republic, Czech Republic) Grant no. 16- 28119a "Analysis of movement disorders for the study of extrapyramidal diseases mechanism using motion capture camera systems", by The Czech Science Foundation (Czech Republic), grant No. 16- 07879S “REM sleep behavior disorder: predicting the risk of neurodegeneration”, and Charles University, Progres Q27.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (26)

Cited by (3)

  • The timed up &amp; go test sit-to-stand transition: Which signals measured by inertial sensors are a viable route for continuous analysis?

    2021, Gait and Posture
    Citation Excerpt :

    Slightly stronger intra-class correlations in the CG than PD were observed in all statistically significant results. This is consistent with a previous study of walking turn signals in PD and a CG [8]. An analysis of the common signal pattern indicated that ωpitch had the most characteristic patterns for StS, which corresponds with intuitive processing of ωpitch in previous studies [3,10].

View full text