Postural Organization of Gait Initiation for Biomechanical Analysis Using Force Platform Recordings

Summary

This paper describes the material and method developed to investigate the postural organization of gait initiation. The method is based on force platform recordings and on the direct principle of mechanics to compute center of gravity and center of pressure kinematics.

Abstract

Gait initiation (GI), the transient phase between orthograde posture and steady-state locomotion, is a functional task and an experimental paradigm that is classically used in the literature to obtain insight into the basic postural mechanisms underlying body motion and balance control. Investigating GI has also contributed to a better understanding of the physiopathology of postural disorders in elderly and neurological participants (e.g., patients with Parkinson's disease). As such, it is recognized to have important clinical implications, especially in terms of fall prevention.

This paper aims to provide scholars, clinicians, and higher education students information on the material and method developed to investigate GI postural organization via a biomechanical approach. The method is based on force platform recordings and the direct principle of mechanics to compute the kinematics of the center of gravity and center of pressure. The interaction between these two virtual points is a key element in this method since it determines the conditions of stability and whole-body progression. The protocol involves the participant initially standing immobile in an upright posture and starting to walk until the end of an at least 5 m track.

It is recommended to vary the GI velocity (slow, spontaneous, fast) and the level of temporal pressure - gait may be initiated as soon as possible after the deliverance of a departure signal (high level of temporal pressure) or when the participant feels ready (low level of temporal pressure). Biomechanical parameters obtained with this method (e.g., duration and amplitude of anticipatory postural adjustments, step length/width, performance, and stability) are defined, and their computation method is detailed. In addition, typical values obtained in healthy young adults are provided. Finally, critical steps, limitations, and significance of the method with respect to the alternative method (motion capture system) are discussed.

Introduction

Gait initiation (GI), the transient phase between orthograde posture and steady-state locomotion, is a functional task and an experimental paradigm that is classically used in the literature to investigate postural control during a complex motor task requiring simultaneous whole-body propulsion and stability¹. Patients with neurological conditions, such as Parkinson's disease², stroke³, progressive supranuclear palsy⁴, and "higher level gait disorders"⁵, are known to have difficulty initiating gait, which exposes them to an increased risk of falling. It is therefore important for both basic and clinical sciences to develop concepts and methods to gain insight into the postural control mechanisms in play during gait initiation, to gain scientific knowledge and a better understanding of the pathophysiology of gait and balance disorders and be able to remediate them through adequate interventions.

The concept of biomechanical organization of gait initiation is described below, and the classical method designed to investigate this organization is detailed in the protocol section. GI can be subdivided into three successive phases: the "anticipatory postural adjustments" (APA) phase corresponding to the dynamic phenomena occurring in the whole body before swing heel-off, the "unloading" phase (between swing heel-off and toe-off), and the "swing" phase that ends at the time of swing foot contacting the support surface. This classical subdivision of the GI process originates from the pioneering studies of Belenkii et al.⁶ and others^7,8, focusing on the coordination between posture and movement during voluntary arm raising to horizontal in the erect posture. In this paradigm, the body segments that are directly involved in the arm raising correspond to the "focal" chain, while the body segments that are interposed between the proximal part of the focal chain and the support surface correspond to the "postural" chain⁹. These authors reported that raising the arm was systematically preceded by dynamic and electromyographical phenomena in the postural chain, which they called "anticipatory postural adjustments". For GI, swing heel-off (or swing toe-off, depending on the authors) is considered as the onset of gait movement¹⁰. Consequently, the dynamic phenomena occurring before this instant correspond to APA, and the swing limb is considered to be a component of the focal chain¹¹. This statement is in agreement with the classical conception of movement biomechanical organization, according to which any motor act must involve a focal and a postural component^12,13.

From a biomechanical point of view, APA associated with GI manifests as a backward and mediolateral (swing leg side-oriented) displacement of the center of pressure, which acts to propel the center of gravity in the opposite direction - forward and toward the stance leg side. The larger the anticipatory backward center of pressure displacement, the higher the motor performance in terms of the forward center of gravity velocity at foot contact^10,14. In addition, by propelling the center of gravity toward the stance leg side, APA contribute to maintain mediolateral stability during the swing phase of GI^1,15,16,17. The current literature stresses that alteration in this anticipatory control of stability is a major source of falls in the elderly¹. Stability during GI has been quantified in the literature with an adaptation of the "margin of stability"¹⁸, a quantity that takes into account both the velocity and the position of the center of gravity within the base of support. In addition to the development of APA, the fall of the center of gravity during the swing phase of GI under the effect of gravity has been reported to be actively braked by the triceps surae of the stance leg. This active braking facilitates stability maintenance after foot contact, allowing a smooth foot landing on the support surface⁴.

The goal of this paper is to provide scholars, clinicians, and higher education students information on the material and method developed in our laboratory to investigate the postural organization of GI via a biomechanical approach. This "global" method (which can also be assimilated to a "kinetic" method for the reasons detailed below) was initiated by Brenière and collaborators^10,19. It is based on the direct principle of mechanics to calculate both the acceleration of the center of gravity, as well as the instantaneous positions of the center of pressure. Each of these points is a global expression specific to the movement.

One is the instantaneous expression of the movements of all body segments related to the purpose of the movement (the center of gravity; e.g., the progression velocity of the body during GI); the other (the center of pressure) is the expression of the support conditions necessary to reach this objective. The instantaneous positions of these two points reflect the posturo-dynamic conditions to be satisfied for gait initiation. The force platform is the appropriate instrument for this model because it allows the direct measurement of the external forces and moments acting at the supporting surface during movement. It also allows the performance of natural movements and requires no special preparation.

Many factors are known to influence the postural organization of GI, including biomechanical, (neuro)physiological, psychological, environmental, and cognitive factors^1,20. This paper focuses on the influence of two factors - velocity of GI and temporal pressure - and provides typical values obtained in healthy young adults.

Protocol

The protocol described below follows the guideline of the human research ethics committee of the Université Paris-Saclay. Participants approved and signed a consent form.

1. Participants

1. Include at least 15 healthy young adult participants in the experiment (aged 20 to 40 years old).
   NOTE: This recommended number of subjects corresponds to what is classically considered in the literature on GI.
2. Exclude participants with walking aids, visual, hearing, or orthopedic problems, identified neurological disorders, dementia, cognitive impairments (i.e., a score < 25 on the Mini Mental State Exam), and a medical history of falling.
3. Ask participants to provide written consent after informing them of the nature and purpose of the experiment.
4. Ensure that the experiment conforms to the standards set by the Declaration of Helsinki.

2. Laboratory preparation

1. Ensure that the force platform is long enough to have the entire swing foot land on it at the end of the first step. If it is not, use two small distance force platforms, with participants standing in the initial posture on the first one and striking their swing foot on the second one placed in front of the first²¹. In both cases, ensure that the force platform(s) is/are embedded in a track at least 5 m long to ensure that steady-state walking is reached.
   NOTE: A force platform that registers the 3D moments and forces is necessary to compute the whole set of experimental variables (see section 5).
   1. As a safety measure, affix a harness to the ceiling and center it to the grand axe of the force platform in case the experiment includes frail patients (e.g., neurological patients).
2. Calibrate the force platform(s). Click on the auto-zero button.
3. Importing the journals
   1. Open Qualisys Track manager.
   2. Choose and open the "Project" folder.
4. Create a patient folder.
   1. Click on Add, then select patients.
   2. Enter labels: Patient ID, First name, Last name, Date of birth, Sex, and Comment if needed.
   3. Click on Add, then select Gait session.
   4. Enter labels: Case ID, Test operator, Comments if needed, Diagnosis, Secondary Diagnosis, Affected Side, Gross Motor Function Classification, Functional Mobility Scale, Height, Weight, Leg length left, Leg length right, Knee width left, Knee width right, Ankle width left, Ankle width right, Sole delta left, Sole delta right, Shoulder offset left , Shoulder offset right, Elbow width left, Elbow width right, Wrist width left, Wrist width right, Hand thickness left, Hand thickness right, and Marker diameter.
   5. Click on Add, then select Markerless session.
   6. Enter labels: Test Condition, Prothesis_Orthosis, External aid, External aid side, Personal aid, Personal aid side, Comments if needed, Test operator, and Event mode (choose multiple force plate).
5. Check Force plate auto-zero.
   1. Select Tools.
   2. Click on Force Plates.
   3. Click on On preview start in the label box "Force plate auto-zero".
6. Ensure that the baseline signals from the force platform (forces and moments) are at zero when it is uncharged.
   1. Click on New or use shortcut Ctrl+N.
   2. Click on Data info Window 1 or use shortcut Ctrl+D.
   3. Click on Display Force Data or use shortcut Ctrl+D.
   4. Click on Force and select Plot.

3. Experimental procedure

1. Ask participants to stand barefoot and immobile on a force platform in their natural upright posture, with the arms hanging loosely against their sides, and their gaze directed to a target at eye level at least 5 m away (Figure 1).
   NOTE: Delineate the position of the feet on the force platform in the initial posture (e.g., with chalk). Carefully check that the participants reposition their feet on these marks after each trial. This point is important since the initial foot position influences the APA features of GI.
2. Determine the participants' preferential starting leg by pushing lightly against the participants' back while in the initial posture with their eyes closed to provoke a step forward.
3. Explain to the participants that the task they are to perform is to initiate gait from the standing posture with the preferred leg, to continue walking to the end of the track, and then to return quietly to the initial standing posture.
   NOTE: If during the experiment gait is not initiated with the identified preferred leg in a given trial, repeat the trial.
4. Explain that gait is to be initiated following two successive signals (acoustic, visual, or tactile): a preparatory signal and a departure signal (see steps 3.6 and 3.7).
5. Explain the instructions on velocity and temporal pressure (see steps 3.8-3.10).
6. Deliver the first (preparatory) signal to the participants. Instruct them to stand immobile and avoid anticipating GI at this first signal.
7. Deliver the second (departure) signal following a random 2-5 s delay after the preparatory signal.
   1. Ensure that the participants are visually immobile before delivering this second signal. Check immobility online with the time plots of the anteroposterior or mediolateral center of pressure displacement
      NOTE: If they are not immobile, detection of APA onset (step 5.1.1) may be difficult.
8. Instruct the participants to either initiate gait i) as soon as possible (i.e., in a reaction time condition), or ii) only once they feel ready (i.e., in a self-initiated condition) following the departure signal.
9. Vary the conditions of "temporal pressure" imposed on GI (i.e., low temporal pressure (self-initiated condition) and high temporal pressure (reaction time condition)).
10. Vary the conditions of GI velocity (slow, spontaneous, fast conditions).
    1. To limit the number of experimental conditions and thus avoid fatigue, instruct the participants to perform only two conditions of GI velocity (e.g., slow and fast) under a low or high temporal pressure condition, or the reverse (i.e., GI at a slow or fast velocity under a high and a low temporal pressure condition).
       NOTE: Repeat the instructions on temporal pressure and GI velocity frequently.
11. Instruct the participants to perform series of 10 successive trials in each experimental condition.
    NOTE: Series of five trials are sufficient for elderly subjects or patients with Parkinson's disease²².
    1. Randomize the conditions of GI velocity and temporal pressure across the participants to avoid order effects.
12. Impose a rest of at least 2 min between successive conditions to avoid the effects of fatigue.
13. In each condition, allow participants to perform two familiarization trials before the recordings.
14. Trigger data acquisition from the force platform a few seconds before the onset of the preparatory signal and stop once the participant has left the force platform.

Figure 1: Experimental setup. The participants initially stand on a force platform (1) embedded in a track at least 5 m long (2), with the gaze directed toward a target at eye level (3). Please click here to view a larger version of this figure.

4. Processing of force platform kinetics recordings

1. Filter data from the force platform using a no-lag low-pass Butterworth order with a 15 Hz cut-off frequency.
   1. Import the file.
   2. Open Visual3D.
   3. Choose and open the file "Project".
   4. Processing
      1. Click on Pipeline or use shortcut F11.
      2. Select Signal Filter.
      3. Select Lowpass_Filter.
      4. Click on Execute.
2. Collect data from the force platform at a rate of 100 Hz.
   1. Click on Pipeline or use shortcut F11.
   2. Select File Save/Export.
   3. Select Export_Data_To_Acsii_File.
   4. Click on Edit.
   5. Enter 100 in the label Number of Points for Normalization.
   6. Click on Execute.
3. Compute the time plots of the instantaneous center of gravity accelerations along the anteroposterior (x''G), mediolateral (y''G), and vertical (z''G) directions from the 3D ground reaction forces obtained with the force platform (see Supplemental Figure S1) using Newton's second law^10,23.
   NOTE: According to Newton's second law, the sum of the external forces applied to a system is equal to the mass of this system (m) multiplied by the acceleration of its center of gravity. Thus, with the GI protocol described in this study, the sole external forces applied to the participants are body weight (BW) and ground reaction forces (R). Equations (1), (2), and (3) can be written:
   x''G = Rx / m     (1)
   y''G = Ry / m     (2)
   z"G = (Rz - BW) / m     (3)
   Where Rx, Ry, Rz are the instantaneous anteroposterior, mediolateral, and vertical components of the vector ground reaction force, respectively. Typical plots of x''G, y''G, and z''G are shown in Figure 2.
4. Compute the 3D time plots of the center of gravity velocity by means of a simple numerical integration of the 3D center of gravity acceleration time plots, using integration constants equal to zero (i.e., 3D initial center of gravity velocity considered as null¹⁰). See Figure 2 for typical time plots of anteroposterior, mediolateral, and vertical velocity of the center of gravity (x'G, y'G, and z'G, respectively).
5. Perform an additional integration of the y'G time plot to obtain the displacement of the center of gravity along the mediolateral direction. Use this quantity to compute the "margin of stability" (see step 5.3.5.2).
6. Compute the mediolateral (yP) and anteroposterior (xP) displacement of the center of pressure from force platform data using equations (4) and (5):
        (4)
        (5)
   Where Mx and My are the instantaneous moments around the anteroposterior and mediolateral directions, respectively; Rx, Ry, and Rz are the instantaneous anteroposterior, mediolateral, and vertical ground reaction forces, respectively; and dz is the distance between the surface of the force platform and its origin (provided by the manufacturer). Typical time plots of xP and yP are shown in Figure 2 (see also Supplemental Figure S2).

5. Experimental variables

NOTE: Each experimental variable described below must be extracted from the experimental time plots obtained for each trial.

1. Detection of the timing events of gait initiation
   1. Onset of APA
      1. Display the time plots of the center of pressure displacement along the mediolateral and anteroposterior directions.
      2. Compute the mean value of the mediolateral and anteroposterior center of pressure time plot during the 250 ms time window preceding the second signal delivered to the participants.
         NOTE: These values correspond to the "baseline values" of these time plots.
      3. Detect the instants following the second signal when the mediolateral and the anteroposterior center of pressure displacement trace deviates 2.5 standard deviations from the baseline value for at least 50 ms.
         NOTE: These two instants correspond to the onset of APA along the mediolateral and anteroposterior directions (t0ML and t0AP, respectively; Figure 2). These two instants may also be identified as the instants when the time plots of the mediolateral and anteroposterior center of gravity acceleration reach 10% of their respective peak value.
      4. Ensure that, in the reaction time condition, the onset of APA ranges between 150 ms and 300 ms after the second (Go) signal. If not, repeat the trial and the instructions on temporal pressure.
         NOTE: If it is less than 150 ms, the participants have anticipated. If it is greater than 300 ms, the participants were not focused on the task.
      5. Ensure that in the self-initiated condition, the onset of APA is greater than 300 ms. If it is not, repeat the trial and the instructions on temporal pressure as the participants may have initiated gait in a reaction time condition.
   2. Swing heel-off time
      1. Display the time plots of the vertical center of gravity velocity and anteroposterior center of pressure displacement.
      2. Identify the instant when the trace of the vertical center of gravity velocity first peaks downward after APA onset as the swing heel-off time²⁴ (Figure 2). Alternatively, identify the instant when the time plot of the anteroposterior center of pressure displacement shows a rapid drop toward the baseline (i.e., toward the toes; Figure 2) or place a foot switch (an inexpensive tool) at the swing heel.
   3. Swing toe-off time
      1. Display the time plots of the mediolateral and anteroposterior center of pressure displacement and of the anteroposterior velocity of the center of gravity.
      2. Identify the instant when the time plot of the mediolateral center of pressure displacement reaches a first (quasi) plateau directed toward the stance foot side as the swing toe-off time (Figure 2). Alternatively, identify the instant following swing heel-off when the time plot of the anteroposterior center of pressure displacement reaches 90% of the maximum backward value, or place a foot switch at the swing toe.
   4. Swing foot contact time
      1. Display the time plots of the anteroposterior center of pressure displacement.
      2. Identify the instant when the anteroposterior center of pressure is abruptly shifted forward (Figure 2) as the swing foot contact time. If this time plot is derived, identify the swing foot contact time as the instant when this derived time plot increases sharply from its baseline level value. Alternatively, place a foot switch at the swing heel to detect this instant.
         NOTE: A method similar to that previously described above for APA detection (based on the computation of a mean baseline level value; step 5.1.1.2) can be used here.
   5. Rear foot-off time
      1. Display the time plot of the mediolateral center of pressure displacement.
      2. Identify the instant when the time plot of the mediolateral center of pressure displacement reaches a second (quasi) plateau, directed in the opposite direction as the first one (step 5.1.3.2; Figure 2), the rear foot-off time²⁵. Alternatively, place a foot switch at the rear to detect this instant.
2. Computation of temporal variables
   1. Compute the delay between the onset of APA (t0ML and t0AP) and the swing heel-off time (tHO) for both the mediolateral and anteroposterior directions, which correspond to the duration of APA along the mediolateral (dAPAML) and the anteroposterior directions (dAPAAP). See equations (6) and (7).
      dAPAML = tHO - t0ML     (6)
      dAPAAP = tHO - t0AP     (7)
   2. Compute the delay between swing toe-off time (tTO) and swing heel-off time (tHO), which corresponds to the "unloading phase" duration (UNLd; Figure 2) using equation (8).
      UNLd = tTO - tHO     (8)
   3. Compute the delay between swing toe-off time (tTO) and swing foot contact (tFC), which corresponds to the "swing phase" duration (SWINGd; Figure 2) using equation (9).
      SWINGd = tFC - tTO    (9)
3. Computation of spatial variables
   1. Initial position of the center of pressure
      1. Display the time plots of the center of pressure displacement along the mediolateral and anteroposterior directions.
      2. Compute the mean values of the mediolateral (yP0) and anteroposterior (xP0) center of pressure positions during the 250 ms time window preceding the second (departure) signal delivered to the participants, which are representative of the center of pressure position in the initial posture (or "baseline" value).
         NOTE: The spatio-temporal features of APA described above are sensitive to the position of the center of pressure in the initial posture²⁶. Hence, it is important to check that any change in the APA features between experimental conditions (e.g., a condition with an obstacle to clear vs. a condition without an obstacle to clear) or between experimental populations (e.g., healthy participants vs. neurological participants) cannot be ascribed to a "simple" change in the center of pressure position in the initial posture, but rather to the factor being investigated.
   2. Amplitude of APA
      1. Display the time plots of the center of pressure displacement and center of gravity velocity along the mediolateral and anteroposterior directions.
      2. Detect the instant when each of these four time plots reaches a maximal value during the APA time window (Figure 2).
      3. Subtract the mean center of pressure baseline value computed in step 5.3.1.2 (i.e., the xP0 and yP0 values) from the maximal center of pressure value detected during the APA time window (for each direction; i.e., compute using equations (10) and (11)).
         xPAPA = xPMAX - xP0     (10)
         yPAPA = yPMAX - yP0     (11)
         Where xPAPA and yPAPA are the amplitude of APA (center of pressure) along the anteroposterior and mediolateral directions, respectively; xPMAX and yPMAX are the maximal anticipatory center of pressure displacement along the anteroposterior and mediolateral directions, respectively.
         ​NOTE: No such baseline subtraction is necessary for center of gravity velocity since it is considered that the participants are initially immobile (the initial center of gravity velocity is therefore null; see step 4.4). The four values obtained are representative of the amplitude of APA (two values per direction).
   3. Step length and step width
      1. Display the time plot of the center of pressure displacement along the anteroposterior direction.
      2. Detect the most backward position of the center of pressure position, xPBACK.
      3. Detect the center of pressure position at the time of rear foot-off, xPRFO (Figure 2 and step 5.1.5).
      4. Compute the spatial difference between these two quantities, which corresponds to step length, L⁴¹, using equation (12).
         L = xPBACK - Xprfo     (12)
      5. Display the time plot of the center of pressure displacement along the mediolateral direction.
      6. Detect the most lateral position of the mediolateral center of pressure position obtained during the first plateau of the time plot, yPSTANCE ("stance", because the center of pressure is located under the stance foot at that time; see Figure 2).
      7. Detect the lateral center of pressure position at the rear foot-off time, yPRFO (Figure 2 and step 5.1.5).
      8. Compute the spatial difference between these two quantities, which corresponds to step width, W²⁵, using equation (13).
         ​W = yPSTANCE - yPRFO     (13)
   4. Performance of gait initiation
      1. Display the time plot of the center of gravity velocity along the anteroposterior direction (Figure 2).
      2. Detect the instant when the participants strike the force platform with the swing foot (step 5.1.4, Figure 2) and note the velocity of the center of gravity at this instant as a criterion of GI performance.
         NOTE: The peak value of this time plot, which is reached a few milliseconds after swing foot contact, can also be considered as a criterion of GI performance. Step length and swing phase duration can also be considered as criteria of GI performance. The longer and the shorter these quantities are, respectively, the better the performance is.
   5. Stability control parameters
      1. For braking index, display the time plot of the center of gravity velocity along the vertical direction. Detect the peak downward center of gravity velocity of the time plot (z'GMIN) and the center of gravity velocity at the swing foot contact time (z'GFC, Figure 2). Compute the difference between these two quantities, termed the "braking index" (BI), as an indicator of stability control, using equation (14).
         BI =      (14)
         NOTE: The BI was introduced by Do and colleagues and provides evidence that the central nervous system anticipates the swing foot strike with the support surface by decreasing the vertical center of gravity velocity during the swing phase of gait initiation^4,5,27. This active braking facilitates stability maintenance after foot strike. The greater the BI, the better the stability control.
      2. For the margin of stability, display the time plots of the center of gravity velocity and displacement along the mediolateral direction. Detect the velocity (y'GFC) and the displacement of the center of gravity (yGFC) at swing foot contact time (Figure 2). Compute the mediolateral component of the margin of stability (MOS) at foot contact using equation (15).
              (15)
         Where BOSmax is the mediolateral boundary of the base of support (BOS) and ω0 is the eigenfrequency of the body, modeled as an inverted pendulum. During GI, participants systematically land on the force platform first with the swing heel, then with the toe. Under such a foot landing strategy, the BOSmax can be estimated with the mediolateral center of pressure position at the time of rear foot-off (step 5.1.5). The eigenfrequency of the body can be computed using equation (16).
              (16)
         Where g = 9.81 m/s² is the gravitational acceleration and l is the length of the inverted pendulum, which corresponds to 57.5% of body height.
         NOTE: The quantity in brackets in equation (15) is termed the "extrapolated center of mass"¹⁸. The condition of the stability at foot contact implies that the extrapolated center of mass is located within the base of support. This condition corresponds to a positive MOS value. If the MOS is negative, corrective postural adjustments are required to recover balance.

Representative Results

Description of representative biomechanical time plots obtained from the force platform during gait initiation
Whatever the level of temporal pressure or the instruction on GI velocity, swing heel-off is systematically preceded by APA. These APA can be characterized by a backward and swing leg side shift of the center of pressure (Figure 2). This anticipatory center of pressure shift promotes the acceleration of the center of gravity in the opposite direction (i.e., forward and to the stance leg side). Along the anteroposterior direction, the velocity of the center of gravity increases progressively to peak shortly after swing foot contact. Along the mediolateral direction, the center of gravity velocity first peaks toward the stance leg side at around swing toe-off, then peaks toward the swing leg side shortly after foot contact. Along the vertical direction, the center of gravity velocity peaks downward at around mid-stance. It then reverses direction and reaches a value close to zero at around foot contact.

Figure 2: Representative biomechanical time plots obtained from the force platform during gait initiation (one single trial) and selected spatio-temporal variables. Gait was initiated quickly in a reaction time condition. X''G, y''G, z''G: acceleration of the center of gravity along the anteroposterior, mediolateral, and vertical directions, respectively. X'G, y'G, z'G: velocity of the center of gravity along the anteroposterior, mediolateral, and vertical directions, respectively. xP, yP: displacement of the center of pressure along the anteroposterior and mediolateral directions, respectively. Timing events. t0ML, t0AP, tHO, tTO, tFC, tRFO: onset of APA along the mediolateral and anteroposterior directions, time of swing heel-off, time of swing toe-off, time of swing foot contact, and time of rear foot-off, respectively. Temporal variables. APA, UNL, SWING: time windows for APA, unloading phase, and swing phase of gait initiation, respectively. Spatial variables. X'GFO, x'GFC, xPMAX, yPMAX, L, W, z'GMIN, z'GFC: anteroposterior velocity of the center of gravity at foot-off and foot-contact, maximal anticipatory center of pressure displacement along the anteroposterior and mediolateral directions, step length, step width, peak downward center of gravity velocity, and vertical center of gravity velocity at swing foot contact time, respectively. Please click here to view a larger version of this figure.

Representative values of experimental variables in young healthy adults: Influence of velocity and temporal pressure

Temporal variables

APA duration
The duration of APA along the anteroposterior and mediolateral directions depends on the velocity of GI but in an opposite way. More specifically, APA duration along the anteroposterior direction increases with GI velocity, with typical values ranging between ~500 ms for slow GI, and ~700 ms for fast GI⁹. By contrast, APA duration along the mediolateral direction decreases with GI velocity. Typical values range between ~700 ms for slow GI and ~500 ms for fast GI²¹.

The duration of anteroposterior and mediolateral APA also depends on the temporal pressure (values provided above are for a self-initiated condition (i.e., a condition with a low temporal pressure level). Studies in the literature typically compare APA duration in a condition with low versus high temporal pressure, when gait is initiated in a fast condition^1,28. Under these conditions, the duration of both anteroposterior and mediolateral APA decreases by approximately 20-30 ms in the reaction time condition compared to the self-initiated condition.

Unloading phase duration
The unloading phase duration depends on the velocity of GI (i.e., it decreases when GI velocity increases). Typical durations range between ~200 ms for slow GI and ~70 ms for fast GI²¹. The unloading phase duration is not sensitive to temporal pressure, at least when gait is initiated in a fast condition²⁹.

Swing phase duration
The swing phase duration depends on the velocity of GI (i.e., it decreases when velocity increases). Typical durations range between ~500 ms for slow GI and ~300 ms for fast GI²¹. By contrast, this duration is not sensitive to temporal pressure, at least when gait is initiated in a fast condition²⁹.

Spatial variables

Amplitude of APA
The amplitude of APA depends on the velocity of GI. More specifically, in a self-initiated condition, the amplitude of APA along the anteroposterior direction increases when the velocity of GI increases⁹. Typical APA values range between ~7 cm and ~0.15 m/s (for the anticipatory center of pressure displacement and the center of gravity velocity, respectively) for slow GI, and ~13 cm and ~0.36 m/s for fast GI. The amplitude of APA along the mediolateral direction, in terms of center of pressure displacement, also increases with the velocity of GI²¹. Typical values range between ~3 cm for slow GI and ~4 cm for fast GI. By contrast, the maximal velocity of the center of gravity reached during APA (mediolateral direction) does not change with the velocity of GI. Typical values are ~0.13 m/s. The amplitude of APA is also sensitive to temporal pressure, at least when gait is initiated quickly^28,29. More specifically, both the anteroposterior and mediolateral components of APA increase with temporal pressure.

Step length and step width
Step length and step width both depend on the velocity of GI but not on temporal pressure. Step length typically reaches ~50 cm and ~90 cm when gait is initiated in a slow and a fast condition, respectively²³. Step width typically reaches ~12 cm and ~14 cm when gait is initiated in a slow and a fast GI condition, respectively⁹.

Performance of gait initiation
The peak of center of gravity velocity typically ranges between ~1 m/s for slow GI and ~2 m/s for fast GI¹⁰. For fast GI, temporal pressure does not affect this performance parameter²⁹, although it may induce a small (~9%) alteration²⁸.

Stability control parameters

Braking index
The BI is sensitive to the velocity of GI. When gait is initiated in a slow condition with a step length less than ~43 cm, the BI is null because there is no need to brake the fall of the center of gravity. The need to brake the center of gravity fall occurs for step lengths greater than 43 cm. A typical value of BI is 0.08 m/s for gait initiated at 1 m/s and with a step length of 55 cm²⁷.

Margin of stability
The MOS is not sensitive to the velocity of GI or to temporal pressure^21,30. Typical MOS values obtained during GI are ~5 cm²¹.

Supplemental Figure S1: Screenshots of the software (Qualisys Track Manager) showing 3D ground reaction forces during gait initiation. Left, the force platform axis, the center of pressure position (corresponding to the application point of the ground reaction force vector), and the ground reaction force vector in the initial posture; right, the time-course of the raw 3D ground reaction forces during gait initiation (one participant, one trial). Green, red, and blue traces represent the ground reaction force along the anteroposterior, mediolateral, and vertical direction, respectively. Ordinate: force amplitude in Newtons. Abscissa: time in ms. The participants initially stood at the left side of the force platform and initiated gait to the right side. Note that the participant left the force platform at time t = 3,200 ms. Please click here to download this File.

Supplemental Figure S2: Screenshots of the software (Qualisys Track Manager) showing the raw center of pressure displacement traces. Left, the force platform axis, the center of pressure position (corresponding to the application point of the ground reaction force vector), and the action force vector exerted by the participant on the force platform in the initial posture; right, the time-course of the raw center of pressure displacement traces (one participant, one trial). Green and red traces represent the center of pressure displacement along the anteroposterior and mediolateral direction, respectively. Ordinate: displacement in millimeters. Abscissa: time in ms. The participants initially stood at the left side of the force platform and initiated gait to the right side. Note that the participant left the force platform at time t = 3,200 ms. Please click here to download this File.

Discussion

The goal of this paper was to provide scholars, clinicians, and higher education students information on the method (the "global" method) used in our laboratory to investigate the biomechanical organization of gait initiation (GI). Critical steps of the protocol, limitations of the method, and alternative methods and applications are discussed below.

A critical step in the protocol is the detection of the timing events of GI (i.e., APA onset, swing heel-off and toe-off, and rear foot-off). The values of both the temporal and the spatial variables related to the organization of GI depend on the correct detection of these events. For each of them, several methods of detection were proposed (these proposed methods are not exhaustive). It is recommended to use the same method throughout data analysis to ensure consistency across trials and experimental conditions and to allow comparison across studies in the literature. However, it is also recommended to use at least two different methods to ensure that the correct timing events are properly detected (only slight differences in the temporal features values are expected across these methods). Furthermore, for each timing event, automatic detection might be applied (e.g., with a MATLAB routine). This routine can be programmed easily through the methods provided in this article. It is strongly recommended to visually check the coherence and "credibility" of the data automatically obtained with these routines. For example, the amplitude of anticipatory center of pressure displacement should not exceed the base of support size. It is expected to be directed backward and toward the swing leg side (except for specific experimental populations); swing toe-off time is expected to occur after swing heel-off; APA onset should not occur sooner than 150 ms before the departure signal or 300 ms afterward (in a reaction time condition). In other words, it is believed that automatic detection alone is not sufficient to properly and "safely" analyze the data; it is essential to have an in-depth knowledge of i) the global time course of the biomechanical plots expected from the force platform and ii) the typical values expected from healthy participants. We believe that, in addition to the ability to program automatic routines, this knowledge is of strong didactic value for higher education students in biomechanics. This is why these two elements are provided in this article.

It is acknowledged that the "global" method has at least two limitations. First, this method does not provide data on the participants' initial posture (i.e., on the relative position of body segments) but provides data on the initial center of pressure and center of gravity position (the relative position of which determines the condition of balance). The same initial center of pressure and center of gravity position could theoretically be reached with an infinite number of postures. In other words, the initial postural conditions under which the participants initiate gait may not be fully controlled with the global method. Visual checking of the participants' initial posture by an experimented researcher or clinician is therefore important if the relative position of the body segments cannot be recorded (e.g., with a camera). Second, the method does not provide information on the contribution of each body segment acceleration (or "local" accelerations) to whole-body center of gravity acceleration. It follows that it is theoretically possible that the acceleration of certain body segments might be compensated by a deceleration of some distant body segments, resulting in a null whole-body center of gravity acceleration during APA³¹. Thus, the use of accelerometers positioned over several body segments (e.g., trunk, hips, legs) might be relevant to complete the force platform data.

An alternative and popular method to compute the whole-body center of gravity during GI is the kinematic method, which is based on recordings using a motion capture system of reflective markers glued to whole-body joint segments. The signals provided by these reflective markers allow the reconstitution of the whole-body skeleton. Based on the size of each body segment thus reconstituted and information provided by anthropometrical tables (e.g., mass and density of bones), the 3D position of the center of gravity of each segment can be computed with the camera software. With these data, it is then possible to compute the 3D position of the whole-body center of gravity. With successive derivation of the position signal, the velocity and acceleration of the whole-body center of gravity can be obtained. To compute the kinematics of the whole-body center of gravity, 53 reflective markers are required³². However, a simplified model with 13 markers was recently proposed by Tisserand et al.³³.

The advantages of the global method (which can be assimilated into a kinetic method since it is based on the recording of forces and moments) over the kinematic method to investigate the postural organization of GI are the following: i) it requires no preparation of the participants, thus saving time, which is particularly important in cases of frail or pathological patients participating in the experiment; ii) it avoids potential errors in the computation of whole-body center of gravity acceleration due to cumulative small errors on marker positioning made by the experimenter, since the global method provides a direct measure of this quantity; iii) the center of pressure position cannot be computed using motion capture systems. The main disadvantage of the global method over the kinematic method was raised above - it does not allow the investigation of posture or segmental coordination.

Now, it is noteworthy that results from the current literature suggest both methods provide equivalent measure of center of gravity kinematics and event timing during locomotor tasks. For example, Langeard et al.³⁴ reported that estimating the center of gravity braking (the "braking index" (BI)) using the global method or the kinematic method during GI was highly reliable. During compensatory stepping reactions, Maki and McIlroy³⁵ reported that the anteroposterior velocity and displacement of the center of gravity computed at foot contact with both methods provided reasonably good agreement in both young healthy adults and the elderly. Similarly, during straight walking on level ground in people with transfemoral amputation, Lansade et al.³⁶ showed that the estimation of center of gravity velocity from force platform data integration was acceptable. Finally, Caderby et al.²⁴ and Yiou et al.²⁵ showed that these two methods provided a similar estimation of swing heel off event and step length/width, respectively, during GI.

The global method was initially applied to the GI paradigm in young healthy adults to obtain basic knowledge on normal postural control during a functional motor task requiring simultaneous whole-body propulsion and stability maintenance¹⁰. It has since been extended widely to investigate many other dynamic whole-body motor tasks, such as lunging in fencing³⁷, jumping³⁸, sit to stand³⁹, and lower limb flexion⁴⁰. It is worth mentioning that the method has also been applied to investigate postural control during the termination of various motor tasks, including single stepping⁴¹ and pointing⁴², and may potentially be applied to investigate gait termination as it has previously been done with the kinematic method⁴³. Finally, the method has also been widely used in patients with neurological conditions and in the elderly to better understand the pathopsychophysiological mechanisms affecting dynamic postural control^2,3,4,5 and, more recently, in patients with Parkinson's disease to test the effectiveness of various non-pharmacological interventions (such as ankle stretching⁴⁴ and functional electrical stimulation³) in enhancing postural control.

In conclusion, this article has presented a detailed method designed to investigate postural control during gait initiation. For each variable, normative values obtained in young healthy adults were provided. The method has a strong biomechanical background, since it is based on the laws of mechanics to compute the kinematics of the center of gravity and the center of pressure. Analysis of the interaction between these two virtual points is a key point of this method, since it determines the conditions of stability and whole-body progression. Because the performance of most of our daily motor tasks (including sports and work) requires safe (stable) whole-body progression, the method is highly appropriate to obtain insight into the posturo-dynamic mechanisms underlying motor efficiency/deficiency in both healthy and pathological populations. It therefore has strong applications in human movement science, sport science, ergonomics, and clinical science.

Materials

Name	Company	Catalog Number	Comments
Force platform(s)	AMTI		One large [120 cm x 60 cm] or two small [60 cm x 40 cm] force platform(s)
Python or Matlab	Python or MathWorks		Programming language for the computation of experimental variables
Qualisys track manage	Qualisys		Software for the synchronization of the force platform(s), the recording and the on-line visualization of raw biomechanical traces (3D forces and moments)
Visual3D	C-Motion Inc		Software for the processing of raw biomechanical traces (low-pass filtering)