Participants
This cohort study comprised stroke patients who underwent gait analysis using a motion analysis system at Seiai Rehabilitation Hospital (Fukuoka, Japan) between 2017 and 2020.
The inclusion criteria were sub-acute stroke patients admitted to the rehabilitation unit, age > 20 years, stroke in the supratentorial area, paresis in one lower limb, ability to understand instructions for performing the gait analysis, and ability to walk 10 m without the assistance of another person or walking aids. The exclusion criteria were patients with bilateral stroke and a history of other neurological or musculoskeletal disorders unrelated to stroke; patients with the PF required to identify the LBF interval was not observed. This research was reviewed and approved by the institutional review board of Seiai Rehabilitation Hospital (approval number: 20–244). All participants gave informed consent prior to participation in the study. This study was conducted in accordance with the Declaration of Helsinki guidelines.
Patient demographics and clinical characteristics
Age, sex, height, weight, time post stroke, stroke type (hemorrhagic or ischemic), paretic side, Fugl-Meyer Assessment (FMA) total score, FMA lower extremity motor score, FMA balance score, trunk control test (TCT) values, and functional accommodation categories (FAC) were investigated at the time of gait measurement.
Fugl-Meyer assessment
The FMA is a 226-point multi-item Likert-type scale developed as an evaluative measure of recovery from hemiplegic stroke. It is divided into 6 domains: upper extremity motor, lower extremity motor, sensory function, balance, joint range of motion, and joint pain. Each domain contains multiple items each scored on a 3-point ordinal scale (0 = cannot perform, 1 = performs partially, 2 = performs fully)26,27. The FMA lower extremity motor score is a subscale measuring lower limb motor recovery. It examines movement, coordination, and reflex action of the hip, knee, and ankle in the supine, sitting, and standing positions. The score range is 0 to 34, with higher scores indicating better lower limb motor performance27. The FMA balance score is a subscale measuring postural control in sitting and standing positions. It measures postural response and retention ability in the sitting, standing, and one-legged standing positions. The score range is 0 to 14, with higher scores indicating better balance ability performance27.
Trunk control test
The TCT has been developed to measure trunk control in stroke patients28. The TCT, besides investigating the maintenance of sitting position, further examines the ability to roll from a supine position towards the paretic side and non-paretic side sides, and supine to sitting position transfer. The score range is 0 to 100, with higher scores indicating better trunk control performance.
Functional accommodation categories
The FAC is an assessment of walking independence29. It has six levels (0 to 5) that are classified according to the walking ability based on the amount of physical support required as follows: nonfunctional ambulatory (FAC 0), continuous manual contact to support the body weight and maintain balance or to assist with coordination (FAC 1), intermittent or continuous light touch to assist with balance or coordination (FAC 2), ambulatory, dependent on supervision (FAC 3), ambulatory, independent, level surface only (FAC 4), and ambulatory, independent (FAC 5).
Experimental conditions
Participants’ task consisted of walking along an 8-m walkway at a self-selected gait speed. The gait cycle of the paretic lower limb was measured at least five times. All participants walked without using any walking aid, such as a cane or orthosis. A therapist accompanied them during gait measurement to prevent falls.
Experimental setup
Fourteen VICON-MX cameras (Vicon Motion System Ltd., Oxford, UK) and six force plates (600 mm × 400 mm; Advanced Mechanical Technology Inc., Watertown, MA, USA) were used for data collection. The sampling frequencies of the cameras and force plates were 100 Hz and 1000 Hz, respectively. Of the 8 m experimental walking path, the measurement section was 6 m and the run-up section was 2 m. Referring to the Helen Hayes Marker Set, 29 reflective markers were attached to the participants’ bodies, including one at the center of the sacrum as an additional marker for the pelvis.
Data collection and analysis
Visual3D analytical software (C-Motion Ltd., Rochelle, IL, USA) was used for data processing. Spatial coordinates of reflective markers and GRF data were processed using low-pass filters of 6 Hz and 18 Hz, respectively. The link segment model consisted of 13 segments (head, trunk, pelvis, upper arms, forearms, thighs, shanks, and feet). The tri-axial joint angle and angular velocity of the lower limbs, the joint moment of the lower limbs, GRF in one walking cycle, and COP were calculated; the data of the joint moment and GRF were normalized using the participants’ body weight.
Based on the vertical component of GRF, one gait cycle was divided into a loading response phase, a single leg support phase, a pre-swing phase, and a swing phase. The threshold of the vertical GRF was set at 1% of the body weight as per a previous study12.
LBF was measured in the posterior component of GRF of the paretic lower limbs after normalizing the gait cycle of the paretic lower limbs as 100%. According to a previous study12, the posterior component of GRF that occurred immediately after the initial contact was defined as FBF, the anterior component that occurred after FBF was defined as PF, and the posterior component that occurred after PF was defined as LBF (Fig. 1). After evaluating LBF, PF impulse and LBF impulse, which are the time integrals of PF and LBF, respectively, were calculated as per previous studies15,30.
The following parameters associated with LBF were selected as analytical variables: TLA, CRP-RMS, the peak value of the ankle plantar flexion moment in the pre-swing phase, and COP forward displacement distance in the paretic stance phase. COP forward displacement distance was normalized by the distance between the heel marker and the second metatarsal head marker on the sagittal plane. The ankle plantar flexion moment was normalized by the participant’s weight. TLA was evaluated as the angle between the perpendicular of the floor from the greater trochanter and the line from the greater trochanter to the distal end of the hind limb. Lewek et al.31 reported that TLA calculated using COP as the distal end had the most similarity to PF. Hence, COP was adopted at the distal end, and the peak value of the TLA of the paretic limb in the single leg support phase was calculated.
CRP-RMS was calculated as per previous studies32,33. First, normalization was performed on the angles and angular velocities in the sagittal plane of the segment using the maximum value of the interval.
$$Angle\!\!:{\theta }_{\begin{array}{c}Ni\end{array}}=\frac{2\times {{[\theta }_{i}-(\theta }_{MAX}+{\theta }_{Min})/2]}{{{\theta }_{MAX}-\theta }_{MIN}}$$
$$Angle Velocity\!\!:{\omega }_{Ni}=\frac{{\omega }_{i}}{{MAX\{MAX(\omega }_{i}), MAX{(-\omega }_{i})\}}$$
The phase angle was then calculated for each segment using inverse trigonometric functions on the normalized angles and angular velocities.
$$Phase\; angle\!\!:{\varphi }_{Ni}={\mathrm{tan}}^{-1}(\frac{{\omega }_{Ni}}{{\theta }_{Ni}})$$
The difference between the phase angles was then calculated as the phase difference, and the RMS during the pre-swing phase was calculated as CRP-RMS.
$${CRP}_{segmentA\_segmentB}={\varphi }_{segmentA}-{\varphi }_{SegmentB}$$
$${CRP}_{\begin{array}{c}segmentA-SegmentB\end{array}}RMS=\sqrt{\frac{1}{N}} \sum_{i}^{N}{({CRP}_{i} )}^{2}$$
Herein, the CRP-RMS values between the pelvic and paretic thigh (CRPPelvis-P_ThighRMS), the paretic thigh and paretic shank (CRPP_Thigh-P_ShankRMS), and paretic shank and paretic foot (CRPP_Shank-P_FootRMS) were calculated as limb coordination of paretic lower limbs. In addition, CRP-RMS values between the paretic and non-paretic thighs (CRPP_Thigh-NP_ThighRMS), the paretic and non-paretic shank (CRPP_Shank-NP_ShankRMS), and the paretic foot and non-paretic foot (CRPP_Foot-NP_FootRMS) were calculated as indicators of cooperativity with the non-paretic side. The following spatiotemporal gait parameters were calculated: walking speed, step length on the paretic limb, step length on the non-paretic limb, loading response phase time of the paretic leg, single leg support phase time, pre-swing phase time, swing phase time, the peak value of the paretic knee flexion angles in the pre-swing and swing phases, and toe clearance during the paretic swing phase. Toe clearance was defined as the distance of the perpendicular line between the head of the second metatarsal bone marker and the floor at the time when the paretic and nonparetic ankle joint markers crossed the sagittal plane during the paretic swing phase. The step length and tip-to-floor distance were normalized to the participants’ height.
Statistical analysis
Statistical analyses were performed using SPSS Statistics Ver.28 (International Business Machines Corp., Armonk, NY, USA). The mean value of the five gait cycles of the paretic limb was used for all gait analysis data. The normality of each parameter was confirmed using the Shapiro–Wilk test; the significance level was set at p < 0.05. First, comparisons of spatiotemporal and kinematic/kinetic parameters of gait in stroke patients with and without LBF were analyzed by either an independent-samples t-test or Mann–Whitney’s U test. Second, regarding the effect of LBF on gait performance, since PF impulse has been reported to affect the spatiotemporal parameters of gait8,9, we calculated partial correlation coefficients between LBF impulse and spatiotemporal parameters with PF impulse as a control variable to remove this effect. In addition, the linear relationships between LBF impulse and the kinetic and kinematic parameters, TLA, peak paretic ankle plantar flexion moment, COP forward displacement distance, and each CRP-RMS, were evaluated using Pearson’s correlation coefficient. Finally, stepwise multiple linear regression analysis was performed with LBF impulse as a response variable and kinetic and kinematic parameters as explanatory variables to identify kinetic and kinematic parameters independently associated with LBF impulse and investigate their relative contributions.