Locomotor circumvention strategies are altered by stroke: II. Postural Coordination

Both post-stroke and healthy participants were recruited through flyers posted at various public locations in the hospital premises of the Jewish Rehabilitation Hospital, in Laval (greater Montreal), Canada. In addition, medical charts were screened for inclusion criteria mentioned below to identify potential participants with stroke. Healthy participants included individuals from the community or volunteers at the Jewish Rehabilitation Hospital. Potential participants were first contacted by secretarial staff not related to the study to give preliminary information about the study and to obtain verbal consent to be contacted by the researchers. After obtaining verbal consent, the first author contacted potential participants to provide detailed information about the experimental protocol and schedule the first experimental session.

Twelve participants with chronic stroke (40–70 years of age) and 12 age-matched healthy controls without any self-reported premorbid conditions that interfered with walking participated in this study, in the period between July 2012 and August 2014. Stroke participants were included if they had a first incidence of supratentorial stroke in the middle cerebral artery territory with an onset of more than 6 months, walking 30 m independently with or without use of a cane, scoring > 27 on the Mini-Mental State Examination [9] obtained from the medical charts, and staging 3/7 or higher on the leg component of the Chedoke McMaster Stroke Assessment (CMSA) [10]). Excluded were those with visuospatial neglect (as screened by Bell’s test [11]) and visual field deficits (reported in medical charts). After screening for exclusion criteria, each participant signed an informed consent form as approved by the research ethics board. Clinical assessment performed in the first session (Fig. 1) included comfortable gait speed using the 10 m walk test [12], motor ability of the lower limbs measured with the help of CMSA (leg and foot impairment stages) [10], cognitive ability using the Montreal Cognitive Assessment (MoCA; [13] and balance confidence measured using the Activities Balance Confidence Scale (ABC; [14, 15]).

The VE, viewed through a helmet-mounted display (HMD; Kaiser Electro- Optics, Carlsbad, CA, USA; FOV: 50° diagonal; screen resolution: 1280 × 1024 pixels), consisted of a room (11 × 8 m) with three red cylindrical obstacles (arranged around an arc of radius 4 m) and a central blue target located at the far end [8]. Participants were asked to walk in the virtual room at a comfortable pace towards the target while avoiding collisions with the obstacles. After advancing 0.5 m, one of the obstacles moved randomly at a speed of 0.75 m/s towards a pre-determined point of intersection (PoI) from head-on, 30° left or right. The PoI was defined as the theoretical point of intersection of the obstacle and participant trajectories (if no avoidance strategy was undertaken and participants walked in a straight line towards the target), and was located at 4 m from the obstacles and the participants’ initial position. In the stationary obstacle condition, the red cylinder stayed still at the PoI. Each obstacle condition (stationary, moving head-on or diagonally left/right) was randomly repeated for 5 trials each over two blocks, with the inclusion of 5 control walking trials with no obstacles presented at the beginning and the end. (Fig. 1).

Reflective markers were placed on 41 pre-determined anatomical locations of the body (Vicon Plug-In-Gait model [16, 17]) and recorded at 120 Hz with a 12-camera motion capture system (512 Workstation, Vicon Motion Systems Ltd. UK). The real-time, 3D positions of three head markers, placed on the HMD, were also used to synchronize motion in the virtual scene with physical movement of the participants in the laboratory space.

After reconstruction of marker positions in 3D, data were filtered using a fourth order, dual - pass Butterworth filter at 6 Hz, and CoM and joint angles (specifically segmental rotations), were extracted from the processed trial data.

Figure 2 illustrates CoM displacements in the anteroposterior (AP) and mediolateral (ML) direction in representative trials from one post-stroke (left paretic) and one healthy control for all obstacle conditions. As the avoidance strategy commenced after initiation of obstacle motion and ended when both obstacle and the participant were at a similar AP position (i.e., at obstacle crossing), further analysis of data was confined between these time periods. Feet marker data were used to generate step width at mid-stance phase at each step for each foot using customized Matlab scripts (Mathworks Inc., MA, USA). For straight walking, step width for each lower limb was simply computed as the distance between the two heel markers in the mid-stance phase. For curved walking as seen in the present study, a point equidistant from the toe and heel marker was identified for each foot as the mid-point. Further, for each foot, a stride vector was identified as the vector representing change in heel marker position between subsequent gait cycles. Step width at mid-stance for each foot was then calculated as the perpendicular distance from the mid-point of the contralateral foot to the stride vector. For instance, left foot width was the perpendicular distance from the mid-point of the right foot to the left stride vector.

The main outcome measures included:

Data reported in this study includes only successful non-collision trials. Stationary vs. head-on and diagonal approaches were analyzed separately. Dependent variables were examined for normality using tests of skewness, kurtosis, Shapiro-Wilk test and Q-Q plots. Data were considered appropriate for analysis by parametric tests if the skewness value was between ±2 [18], if the Shapiro-Wilk test was non-significant and if the spread of data on the Q-Q plot did not deviate far from the expected normal slope. All data used for analysis in this study met these parameters. Between group differences (stroke vs. healthy) with respect to participants’ age and clinical assessments (gait speed, MoCA and ABC scores) were analyzed using unpaired t-tests. The proportion of side step vs. cross step strategy for each group and obstacle condition was analyzed separately using the paired t-test. Separate two-way (2 × 2) repeated measures mixed model analyses were used to determine the impact of group (stroke vs. control) and obstacle approach (stationary vs. head-on) on head, thorax, pelvis, L/P foot, R/NP foot yaw and thorax roll amplitudes, respectively. Similarly, separate two-way repeated measures mixed model analyses were used to determine the effects of group and obstacle approach (R/NP vs. L/P) on joint amplitudes for the above mentioned segments except thorax roll (as thorax roll was not significantly different for both approaches as compared to the no-obstacle trials). Two-way repeated measures mixed model analyses were used to determine impact of group and joint segment on reorientation onsets for stationary and head-on approaches separately (as onset of segment reorientation was identified differently for these conditions). A three-way 2 × 2 × 5 repeated measures mixed model analysis was performed to detect differences in reorientation onsets among groups (stroke vs. control), obstacle approach (L/P vs. R/NP) and joint segments (head, thorax, pelvis, L/P foot and R/NP foot rotations) for diagonal obstacle approaches. Bonferroni adjustments were used for all repeated measures analyses, with significance set at p < 0.05. All statistical analyses were performed using the IBM SPSS Statistics 20 software (IBM Corporation, NY, USA).