Quantifying Gait Asymmetry in Stroke Patients: A Statistical Parametric Mapping (SPM) Approach

Introduction

Stroke is the most prevalent cerebrovascular disease and is a major contributor to adult disability. With aging populations, the number of stroke cases is steadily rising, and projections estimate that by 2030, there will be 70 million stroke patients globally and 12 million stroke-related deaths [1]. Stroke is a major cause of disability, as it is accompanied by muscle weakness and decreased physical coordination and ability to maintain balance [2]. In particular, hemiplegic (ie, one-sided) patients present with asymmetrical gait after stroke due to damage to cerebrovascular and neural circuits [3,4]. This asymmetrical gait pattern causes quality of life impairments (eg, reduced walking speed, asymmetrical lower-limb joint angles and step length) and results in secondary problems such as falls. Restoring a more symmetrical (ie, functional) gait is the main goal of rehabilitation programs following a stroke. However, it has been reported that only 50% of patients are able to walk independently without assistance after rehabilitation and only 7% of patients recover to the point where they can carry out activities of daily life without problems [2,5,6]. Despite rehabilitation efforts, the most effective variables for evaluating and promoting gait recovery after a stroke have yet to be established [2,6]. Differences in muscle strength between the affected side (AS), also referred to as the paretic side, and the unaffected side (UAS) of stroke patients cause asymmetrical gait patterns, such as a short-stance phase and a relatively long swing phase on AS [7]. It has been reported that such kinematic characteristics of asymmetric gait occur due to a decrease in hip joint flexion and knee joint flexion on AS [8,9]. However, to evaluate gait asymmetry in stroke patients, common measurements utilize spatiotemporal variables that are relatively simple and easy to calculate (eg, step length, step time, stance time, and swing time) when comparing AS and UAS [10]. Previous studies on gait asymmetry in stroke patients using spatiotemporal variables have shown conflicting results on the causes of asymmetry. For example, the asymmetry of step length and step time between AS and UAS increases [11], but another study showed that the asymmetry of step length decreases in the gait of stroke patients [6]. Due to the discrepancies in research findings, there has been a shift toward incorporating biomechanical variables alongside spatiotemporal variables to better understand the asymmetrical gait of stroke patients [2]. Recently, there have been increasing numbers of studies analyzing gait asymmetry in stroke patients using kinematic data (eg, joint angle and angular velocity) to examine for potential differences [4]. This provides a more in-depth comparison between AS and UAS than just the spatiotemporal variables, and it offers a variety of different applications [10]. Additionally, because the gait is made of complex movements of multiple joints, equipment such as the VICON 3D motion capture system must be used to collect accurate biomechanical data [12].

Following appropriate data reduction techniques (eg, filtering of electrical signals), interpretation of the results is required. Representative methods include indexing methods such as the Gait Deviation Index (GDI) and Gillette Gait Index (GGI), and methods using statistical analysis such as the paired samples t test [10]. These methods use discrete (ie, single) time points to simply condense the results of gait, which involve complex mechanisms, and reduce them to a representative value. As a result, researchers are unable to assess movements throughout the duration of the gait cycle [13]. For a deeper understanding of gait patterns present in stroke patients, it is necessary to examine the asymmetry during a broader portion of the gait cycle, rather than a single representative value [10]. Although the movement data for stroke patients is continuously collected, previous studies have primarily analyzed gait asymmetry using discrete variables, thereby limiting insights into the continuous dynamics of gait [6,14,15]. Therefore, there is a lack of research that identifies gait asymmetry in stroke patients using continuous biomechanical variables.

Statistical Parametric Mapping (SPM) is a statistical technique that has been widely used in medicine for analyzing functional magnetic resonance imaging (fMRI) data. Additionally, this method can compare gait patterns throughout the entire gait cycle, making it suitable for analyzing asymmetry between AS and UAS that occurs in specific sections, rather than in a discrete representative value (eg, max or min values). Also, SPM has the advantage of minimizing statistical errors and identifying parts that may be overlooked by discrete variables [16]. Therefore, it can accurately identify areas where gait asymmetry has deteriorated in a condensed portion of the gait cycle [16,17]. For example, center of mass (CoM) stability is crucial for energy-efficient gait, as it reduces energy costs during walking. Stroke patients often exhibit greater variability in CoM due to gait asymmetry, which leads to higher energy expenditure and reduced stability. This increased variability requires additional energy to transform potential energy into kinetic energy, further exacerbating the energy demands of walking. Excessive CoM oscillation disrupts smooth and efficient movement, making it challenging for stroke patients to conserve energy. These subtle differences are difficult to detect using discrete variables; however, SPM analysis of continuous variables effectively identifies sections with significant CoM variability, offering a clearer understanding of the specific phases where gait asymmetry occurs [16].

The VICON motion capture system was chosen because 3D motion analysis is widely recognized as the most valid and reliable method for quantifying human movement. Its ability to provide detailed, accurate measurements of joint kinematics makes it an ideal tool for assessing gait asymmetry in post-stroke patients. This high level of validity and reliability ensures that the findings from this study are both credible and replicable [12]. Therefore, the purpose of this study was to analyze the asymmetry of the hip, knee, and ankle joint angles between AS and UAS during gait in stroke patients using SPM to solve problems that may arise from analyzing discrete variables.

Material and Methods

PARTICIPANTS RECRUITMENT:

The total study population consisted of 23 participants, who were recruited from a local welfare center. Participants were selected based on inclusion and exclusion criteria. The inclusion criteria were a Mini-Mental State Examination (MMSE-K) score of 24 or higher, the ability to walk more than 10 meters, and a history of more than 2 years after stroke. The exclusion criteria included the absence of heart disease and visual impairments. Additionally, 3 male participants were excluded to reduce sex-related bias. Subsequently, participants were prioritized based on their MMSE-K scores, with those scoring higher being selected first (Table 1). Ultimately, 20 stroke patients were included in the study. Prior to the study, the purpose of the study was explained to all participants in detail, and informed consent was obtained. Participants were reminded that they could withdraw from the study at any time. The human participant research protocol was approved by the Institutional Review Board of Silla University (1041449-202403-HR-001) and the study procedures were explained to the participants prior to data collection.

EXPERIMENTAL SET-UP AND DATA COLLECTION:

A 250-Hz 10-camera VICON motion capture system (Centennial, CO, USA) was used to capture the gait motion trajectories of the reflective markers attached to the participant’s lower body. Camera calibration was performed prior to data collection. The +Y-axis of the global reference frame (ie, laboratory standard) was defined as the walking direction, and the vertical direction (ie, from the floor to the ceiling) was defined as the +Z-axis. Therefore, the +X-axis of the global reference system was defined in the medio-lateral direction. A total of 23 reflective markers and 4 clusters were attached to each joint and segment of the participant’s lower body: greater trochanter, sacrum, iliac crest of pelvis, anterior/posterior superior iliac spine of pelvis, medial/lateral epicondyle of femur, medial/lateral malleolus of tibia and fibular, heel, toe, 1st and 5th head metatarsophalangeal joint of foot, and clusters on the thigh and shank (Figure 1). VICON Nexus (version 1.8) was used in marker labelling. After the participants sufficiently warmed up, they were instructed to walk 5 times at their preferred speed, and video data capturing 5 strides were collected for each participant. It took approximately 1 hour for each participant to complete all procedures (explaining the consent form, attaching markers to the participant’s body, and conducting data collection).

DATA PROCESSING AND DATA ANALYSIS:

Captured trial motion data stored in C3D files were imported into Kwon3D Motion Analysis Suite (version XP, Visol, Seoul, Korea) for data processing. The markers’ raw coordinates were digitally filtered using a Butterworth 4^th-order zero-phase lag low-pass filter. The cutoff frequency was set to 6 Hz. The orientation angles of the hip, knee, and ankle joints of AS and UAS were analyzed in the sagittal, frontal, and transverse planes. The orientation angle relative to its respective proximal segment was computed for the analysis. The orientation angle was calculated based on the X-Y-Z Cardan rotation sequence.

The gait cycle was divided into a stance phase and a swing phase based on whether the reference leg was in contact with the ground, and each phase was normalized to 100% time for analysis (Figure 2). The asymmetry index was calculated with reference to a previous study [18].

STATISTICAL ANALYSES:

The dependent variables used in the analysis were the orientation angles, in each plane, of lower-extremity joints for the AS and UAS. The average data of the 5 trials in all data analyses were used for statistical analysis. After calculating orientation angles of each joint, Python 3.9 (Python Software Foundation, USA) was used to analyze the continuous entire section (waveform). A paired t test was performed using the provided 1D-SPM (one-dimensional statistical parametric mapping) technique made available as open-source code (www.spm1d.org) [19]. Statistical normality was tested using the built-in function ‘spm1d.stats.normality.ttest’. In biomechanical research, traditional methods using discrete variable analysis simplify continuous kinetic and kinematic data, while 1D-SPM is more suitable for evaluating entire movement curves.

This analysis method can compare gait patterns throughout the whole gait cycle, making it suitable for analyzing asymmetry between the AS and UAS that occurs in specific sections, rather than relying on discrete representative values (eg, maximum or minimum values). The level of statistical significance (α) was set at 0.05.

Results

There were differences observed among lower-extremity joint angles between AS and UAS during gait in stroke patients using SPM, which allows for the comparison of gait patterns throughout the whole gait cycle. For the hip joint angle in the sagittal plane, asymmetry between AS and UAS was statistically significant in the flexed position (0–39%) during the stance phase (Table 2, Figure 3A) and the flexed position (40–100%) during the swing phase (Table 3, Figure 4A). In addition, asymmetry was statistically significant at the flexed position (0–79%) during the swing phase and flexed position (98–100%) during the stance phase in the knee joint angle in the sagittal plane (Table 2, Figure 3D and Table 3, Figure 4D), respectively (P<.05).

Discussion

We found that the SPM allowed for a detailed comparative analysis of the kinematic differences between the AS and UAS during specific phases of gait in stroke patients. Significant differences were identified in the sagittal plane at the hip and knee joints. At the hip joint in the sagittal plane, significant differences were observed from heel contact to mid-stance (0–39%) during the stance phase and from mid-swing to terminal swing (40–100%) during the swing phase. Additionally, at the knee joint in the sagittal plane, significant differences were identified from initial-swing to mid-swing (0–79%) during the swing phase and during toe-off (98–100%) in the stance phase.

The gait cycle is divided into 2 components: the stance phase, in which the reference leg is in contact with the ground, and the swing phase, in which the same leg is in the air during the contralateral leg in stance. The stance phase is divided into heel contact, mid-stance, and propulsive phases. Gait is a continuously connected movement [20], and if a problem occurs in a specific section, compensation may be necessary thereafter. Therefore, emphasis should be on preventing and alleviating abnormal gait patterns. Asymmetric gait patterns for stroke patients result from excessive use of UAS to compensate for slow movement and difficulty shifting weight to the AS, leading to a left/right asymmetric gait [21,22]. Asymmetric gait is caused by the AS shifting the whole-body COM to UAS as quickly as possible due to instability of the lower extremities. This causes AS to perform an equine gait pattern (ie, the toes of the foot hit the ground before the heels) without sufficient flexion of the hip joint during the stance phase [23,24]. A previous study argued that the lack of hip joint flexion movement negatively impacts the distal lower extremity (eg, the foot) and leads to circular gait to maintain toe-heel clearance. However, those findings were derived using maximum flexion values and were simply used to estimate correlations. In contrast, the present study utilized the SPM technique with continuous variables, enabling a precise identification of the specific phases of hip joint dysfunction during gait [25]. The results of our study showed that the hip flexion angle of AS was significantly smaller than that of UAS from the heel contact to mid-stance (0~39%) in the stance phase (Figure 3A). This may be inappropriate for converting GRF into elastic energy through heel contact during the stance phase, storing it in the Achilles tendon and then using it as forward propulsion. In addition, in asymmetric gait, the swing phase is shortened to move COM as quickly as possible to UAS due to the instability of AS lower extremities, resulting in a decreased propulsion and a shortened step length [23,24].

Previous studies have reported that in stroke patients, the hip flexion angle of AS was smaller than that of UAS during the swing phase [10]. However, these studies used discrete variables and Normalized Cross Correlation technique to compare the difference in joint angles between AS and UAS, which did not allow for identifying the specific section of the swing phase where a significant difference occurred. We also found that the hip flexion angle of AS was significantly smaller than that of UAS in most sections (40~100%) of the swing phase, except for the initial part (Figure 4A). Specifically, the results showed insufficient flexion of the hip joint from the mid-swing to the terminal swing phase in AS, which likely contributes to a shortened step length due to inadequate distance covered. This suggests that restoring proper hip flexion in the swing phase may be an important factor for improving gait smoothness and efficiency in rehabilitation for stroke patients. Furthermore, the results indicate that rehabilitation programs should not only focus on improving ankle dorsiflexion, as commonly emphasized in previous studies, but also focus on improving hip and knee joint flexion angles. These joint angles play a critical role in reducing gait asymmetry and improving overall gait efficiency. Targeting these areas may enhance the stability of the CoM, minimize compensatory mechanisms, and ultimately reduce energy expenditure during gait [16,25].

The human body is a connected chain; if knee movement is restricted, compensation occurs in other joints [26]. A typical after-effect of hemiplegic strokes is decreased joint range of motion in the lower extremities on the AS. Due to the asymmetric movement of both lower extremities, there is an increase in the COM variability, which can reduce the energy efficiency of walking [27]. Additionally, according to a previous study utilizing discrete variables, insufficient clearance during walking in stroke patients was attributed to inadequate flexion of the knee and hip joints, as well as insufficient ankle dorsiflexion [28]. In this study, the SPM analysis using continuous variables revealed that knee joint flexion was the most insufficient among these factors. The AS flexion angle of the knee joint was reduced compared to the UAS in the terminal toe-off section (98–100%) in the stance leg and in the initial to mid-swing (0–79%) section in the swing leg (Figures 3D, 4D). Immediately after toe pushing in the stance leg, sufficient knee flexion is necessary to reduce energy consumption and secure toe-heel clearance from the ground by reducing the moment of inertia of the lower extremity from the initial to middle swing in the swing phase. The section where a significant difference occurred in this study requires sufficient knee joint flexion for an efficient gait, but it is believed to be due to the stiff-knee gait characteristic of AS, a common ailment of a stroke. A previous study also reported that the knee joint flexion angle of AS was smaller than that of UAS during the swing phase due to the stiff-knee gait [10]. Stiff-knee gait makes it difficult to activate the muscles properly due to the rigidness of the rectus femoris, resulting in compensation issues such as reduced knee joint bending, increased energy consumption, and circular walking [29,30]. It has been reported that due to the rigidity of the rectus femoris, the knee extension moment is increased, which consequently leads to a significant reduction in knee flexion velocity during toe-off [28]. Additionally, when applying strong pressure against the ground during the toe-off of the stance phase, flexion of the knee joint naturally occurs during the swing phase. A previous study compared GRF during gait between stroke patients and healthy individuals and found that GRF was significantly lower on the AS during the toe-off phase compared to healthy individuals, while there was no significant difference in the UAS during the same phase. Therefore, as indicated by the results of this study, the higher knee flexion angle on the UAS may be attributed to the higher GRF on this side compared to the AS [31], suggesting that the stiff-knee gait does not generate sufficient force in the toe-off section, limiting knee flexion motion.

Previous research has shown that stroke patients have decreased dorsiflexion strength and joint range of motion in the AS ankle joint [32]. To solve this problem, it has been recommended to wear a passive ankle-foot orthosis, a device that can assist dorsiflexion, on the AS ankle join in stroke patients [32].

One study [25] conducted gait analysis on stroke patients and reported that the ankle dorsiflexion angle is highly associated with the hip flexion angle and walking speed, but no statistically significant differences were observed in the maximum dorsiflexion angle. The results of the present study, using SPM, showed no difference in the asymmetry of ankle joint movements between AS and UAS in both the stance phase and swing phase in the sagittal plane (Figures 3G, 4G). This outcome can be explained by previous studies using a discrete variable that represents the maximum flexion/extension angle or joint range of motion rather than the overall gait pattern. Thus, the previous studies are considered to have limitations because they solely presented summarized findings. Previous studies have reported that when circular gait occurs as a compensation for stiff-knee gait, abduction and rotation of the hip joint are induced, and differences in kinematic variables between AS and UAS can be confirmed in the frontal and transverse plane movements [28]. However, in the present study, no statistically significant differences were identified in the frontal and transverse planes (Figures 3B, 3C, 3E–3I, 4B, 4C, 4E–4I). This lack of significant differences is believed to be due to the omission of upper body movements in this study. Among the gait characteristics of stroke patients, most of the body weight is supported on UAS due to limited weight transfer to AS, which can cause the upper body to tilt [28]. The upper body tilting toward UAS helps secure toe-heel clearance between the ground and AS and enables a straight gait rather than a circular gait. In other words, the presence or absence of circular gait on AS is largely influenced by the upper body tilt of stroke patients. However, since our study did not consider such a compensatory effect, it is believed that significant asymmetry between AS and UAS could not be confirmed in the frontal and transverse planes. Therefore, a follow-up study should investigate the control of upper-body movements or to investigate the correlation with lower-body movements based on the categorization of upper-body movements.

The main limitation of this study is the small sample size. A total of 20 stroke patients were recruited for this study and it did not have a control group. While the generalizability is somewhat limited, the findings of the study still offer insight into the asymmetry of the hip, knee, and ankle joint angles between AS and UAS during walking in stroke patients. Also, in this study, differences in functional performance and individual walking patterns based on disease onset were not considered. As a result, a wide range of standard deviation was observed. The absence of a matched control group is another limitation. In future studies, we plan to include a matched control group to enhance the reliability and generalizability of the findings. It is also necessary to control upper-body angles, such as using a harness to stabilize the upper body, which could help minimize the influence of compensatory movements and provide a clearer assessment of lower-limb asymmetry.

Conclusions

The purpose of this study was to analyze the asymmetry of the hip, knee, and ankle joint angles between AS and UAS during gait in stroke patients, using SPM to solve problems that may arise from analyzing discrete variables. In stroke patients, the range of motions of the hip and knee joints on AS were smaller than those of UAS in both stance and swing phase in the sagittal plane during walking. This difference in joint motion between the 2 lower extremities can cause various functional and pathological problems (eg, COM variability and increased load on UAS). In this study, stroke patients showed limited flexion movements in the following sections in AS compared to UAS: first, at the hip joint, heel contact and mid-stance (0% to 39%) in the stance phase, and all sections and except the initial swing (39% to 100%) in the swing phase; and second, at the knee joint, last section of the toe-off (98% to 100%) in the stance phase and from initial to mid-swing (0% to 79%) in the swing phase. Therefore, to develop a successful rehabilitation program for stroke patients, restoring not only ankle dorsiflexion, as predominantly emphasized in prior studies, but also hip and knee joint flexion angles, is a vital element of the rehabilitation process.