1 Introduction

Approximately 1.4 million traumatic brain injuries (TBIs) occur annually in the United States, more than 85 % of which are considered as mild TBIs, also known as concussions [1]. Since most concussions result from collisions in sports, such as American football, rugby, ice hockey, skiing, and boxing, sports concussions are of particular concern [2]. Concussions have been found to be associated with cognitive (i.e., attention, memory, and executive function) deficits [3] and motor (i.e., balance and coordination) dysfunction [4, 5]. These symptoms can persist for up to 1 year, which may affect a person’s daily activities or work performance [6, 7]. Moreover, recurrent concussions are frequently reported within several months after the first concussion [8] and may lead to permanent brain damage or death [9]. Therefore, it is important to accurately detect functional deficits and monitor recovery following the initial concussion so that preventive interventions can be implemented to avoid subsequent injuries.

Since brain regions activated while cognitive functions are performed can also be engaged during motor tasks [10, 11], dual-task gait paradigms have been used to investigate the effect of concussion on cognitive and motor performance [12–15]. Greater toe-obstacle clearances were observed in individuals with concussion while simultaneously performing a visual Stroop task during gait [14]. Previous studies also demonstrated that gait and balance deviations in concussed individuals could be reliably detected with the use of level ground walking with concurrent cognitive tasks [12, 13, 15]. Individuals following mild TBI exhibited a significantly reduced anterior velocity at the center of mass (COM) and a shortened separation distance between the COM and the center of pressure (COP) [12, 13, 15], with the former change lasting up to 28 days post injury [13]. In frontal plane control, a greater medio-lateral COM displacement or COM-COP separation distance was reported [13], with the medio-lateral COM displacement tending to normalize by 14 days post injury [13].

While the effect of concussion on the control of whole body COM motion and its interaction with the COP have been examined, how concussion alters the coordination of the lower extremity joints remains unclear. Since movement of the whole body COM is determined by motions of all body segments [16], measuring changes in the control of joint motion could provide insight into the causation of gait deviations following a TBI. While the examination of kinematic data on each individual joint could be useful in this regard, information on inter-joint coordination is considered to be more informative for determining how the neuro-musculoskeletal system organizes redundant degrees of freedom (DOFs) of the joints to achieve a smooth, efficient, and accurate functional movement [17]. Studies on the pattern and variability of inter-joint coordination using the continuous relative phase (CRP) were reported for young and older adults as well as patients with knee osteoarthritis [18, 19]. Older individuals exhibited similar inter-joint coordination patterns but with increased variability, compared to those of young adults, for the leading limb crossing an obstacle [19]. Bilateral knee osteoarthritis, a localized joint disease, was not found to alter the pattern and variability of the inter-joint coordination of older individuals [18]. It seems that patterns of inter-joint coordination are not altered due to ageing or individual peripheral joint disease. The effect of a brain injury, such as concussion, on the pattern and variability of the inter-joint coordination of the locomotor system remains unknown.

The present study thus investigates the effect of concussion on the inter-joint coordination of the lower extremities during level walking with and without a concurrent cognitive task. It was hypothesized that the gait temporal–spatial variables, patterns, and/or variability of the inter-joint coordination would be changed following a concussion, especially when the individual walked while simultaneously performing a cognitive task.

2 Methods

Fifteen young adults (6 females, 9 males, age: 21.3 ± 3.3 years, height: 179.6 ± 11.4 cm, mass: 88.5 ± 18.2 kg) who had suffered a concussion (CONC) participated in this study. All concussed subjects were diagnosed as having a grade II concussion as defined by the American Academy of Neurology [20], which entails transient confusion and concussion symptoms or mental status changes lasting longer than 15 min, but no loss of consciousness. They were initially identified by certified athletic trainers or attending medical doctors in the university intercollegiate athletic program or the student health center, and were referred for testing as soon as possible following concussion (within 48 h). Individuals were excluded if they had suffered a prior concussion within the previous year or had pre-existing injury or surgery which could affect their gait and/or cognitive abilities. A group of 15 control subjects (NORM), who were free of any neuro-musculoskeletal pathologies, were matched to the CONC by gender, age (21.2 ± 3.4 years), body mass (84.8 ± 22.1 kg), height (179.1 ± 9.9 cm), type of sports participation, and level of education. If any control subjects had symptoms that are commonly reported after a concussion, such as vision problems, nausea, or headache, they were excluded from participation. Experimental procedures were approved by the Research Ethics Committee and were explained to all subjects prior to testing. Written informed consent was obtained from all participating subjects.

The post-concussion symptom scale of immediate post-concussion assessment and cognitive testing (ImPACT), a commonly used neuropsychological test in sports settings, was employed for all the participants to assess the symptoms associated with concussion, such as balance problems and difficulty concentrating. The subjects’ current status were recorded on a scale from zero (not experiencing this symptom currently) to six (suffering greatly from this symptom). The composite symptom score was then calculated as the sum of all the symptoms and their severity.

The gait protocol consisted of two conditions: (1) walking with undivided attention (single-task) and (2) walking while simultaneously performing a cognitive task (dual-task). Each subject walked at a self-selected pace along an 8-m walkway. Several practice trials were allowed so that the subject could walk comfortably with markers along the walkway. Twenty-five retro-reflective markers were placed bilaterally on bony landmarks of the body to track the motion of body segments [12]. Marker trajectories were collected with an eight-camera motion analysis system (Motion Analysis, Santa Rosa, CA, USA) at a sampling rate of 60 Hz. The dual-task walking required the subject to continuously answer the given cognitive questions during the entire walking trial. The cognitive questions consisted of spelling 5-letter words in reverse, subtraction by sevens, or reciting the months of the year in reverse order, which are commonly used in clinical mental status examination to assess cognitive function [21]. The concurrent cognitive task provided interference by using up the available attention during gait. The type of dual-task was randomly selected for each walking trial. Data from five trials were collected for each testing condition.

Marker trajectory data were low-pass-filtered using a fourth-order Butterworth filter with the cut-off frequency set at 8 Hz. Markers placed on each body segment were used to define the local coordinate system. Rotation matrices of adjacent segments were used to calculate the joint angular displacement. OrthoTrak software (Motion Analysis Corp., Santa Rosa, CA, USA) was used to calculate joint angles and gait temporal-distance variables. With joint angles in the sagittal plane, angular velocities were estimated for each joint using the generalized cross-validation spline (GCVSPL) algorithm [22] for a gait cycle. Angular displacements were normalized such that the range of angular positions during movement were confined between −1 and 1, with the midpoint located at zero. Angular velocity values were normalized by the maximum absolute velocity of each trial during the movement, maintaining the absolute zero velocity as the normalized zero velocity [23]. Since the interjoint coordination study focused on the coordination patterns of two joints, such normalization procedures were used as they have been reported to minimize the influence of different movement amplitudes and frequencies of different joints [24]. Phase plots with the normalized angular velocities (x′) against the normalized angular displacements (x) for each joint were then generated for the gait cycle, and the phase angle (φ) was calculated as φ = tan−1(x′/x). The phase angle has been used to characterize biological systems [25, 26]. The CRP values, representing the coordination between two adjacent joints, were then calculated by subtracting the phase angle of the distal joint from that of the proximal, φhip–knee and φknee–ankle, respectively [27]. For each inter-joint relationship, CRP curves from five trials of each subject were ensemble-averaged to determine the general patterns of inter-joint coordination. The deviation phase (DP) was then calculated as the average value of all standard deviations calculated for each point on the ensemble curve over the stance and swing periods, respectively [28]. A lower DP value indicates a smaller variability in the coordination between the two joints [28].

The percentage of correctness in answering the cognitive task during the dual-task (i.e., the ratio between numbers of correct answers and all answers attempted) was calculated to quantify the performance on cognitive tasks. The independent t test was used to detect the group differences in cognitive task performance. All calculated variables were determined to be normally distributed by the Shapiro–Wilk test before being prepared for the repeated measures (2 × 2) mixed design analysis of variance (ANOVA), which was used to determine whether differences existed between groups and tasks for the gait temporal-distance and DP variables. All the significance levels (α) were set at 0.05. Coefficients of multiple correlation (CMC) of the phase angle and CRP curves were calculated to assess the similarity of the curve patterns between each concussed subject and his/her matched control [29]. A CMC value close to 1 indicates similar patterns. All statistical analyses were conducted using SPSS (Version 15.0, Chicago, IL, USA).

3 Results

Concussed subjects demonstrated a significantly higher composite symptom ImPACT score than that of control subjects (CONC: 27.93 ± 25.07, NORM: 1.57 ± 3.18, p = 0.002) at the time of study participation. Percentages of correctness in cognitive task performance during the dual-task did not differ between the CONC (93.3 % ± 7.6) and NORM groups (92.1 % ± 8.6) (t = 0.34, p = 0.74). However, the CONC group adopted a significantly slower gait velocity than that of the NORM group during both single- and dual-task walking (Table 1). During the dual-task walking, both groups adopted a significantly slower gait velocity and shorter stride length (Table 1) compared to those for single-task walking. There were no significant group or task effects for the step width (Table 1).

Table 1 Means and standard deviations of temporal–spatial gait variables for concussion (CONC) and normal (NORM) groups
Full size table

An examination of the phase angle patterns of each individual joint indicates that the levels of similarity between the CONC and NORM groups, in terms of the CMC values, are different for all three joints. For both single- and dual-task walking, the greatest CMC values (0.86 and 0.88) were observed at the hip, gradually declining for the knee (0.74 and 0.66) and ankle (0.65 and 0.62) joints.

For inter-joint coordination, similar hip–knee and knee–ankle CRP patterns were found between the single- and dual-task walking (Fig. 1). For both conditions, the hip–knee CRP showed a w-shape with two negative peaks during the stance phase and close to 0° during the swing phase (Fig. 1a, b). Such w-shape patterns were also found in the knee–ankle CRP but distributed among the entire gait cycle (Fig. 1c, d)., The CMC values between CONC and NORM groups were greater than 0.8 for both the hip–knee and knee–ankle CRP curves (hip–knee: single-task: 0.83, dual-task: 0.81; knee–ankle: single-task: 0.81, dual-task: 0.81), indicating that both groups coordinated their joints in similar patterns during both walking conditions.

Fig. 1
figure 1

Ensemble-averaged CRP curves of hip–knee and knee–ankle coordination of the lower limbs for the concussion (solid lines) and normal (dashed lines) groups during a, c single-task and b, d dual-task walking. Vertical lines indicate the end of the stance phases

Full size image

An examination of the variability of CRP curves using the DP values indicates no significant interactions between the group and condition effects (Table 2). The DP values of the swing phase hip–knee CRP curves of the CONC group were significantly greater than those for the NORM group for both conditions. There were no significant group or condition effects on DP values of the swing phase knee–ankle CRP curves, nor were there any on the stance phase hip–knee and knee–ankle curves.

Table 2 Means and standard deviations of DP values for concussion (CONC) and normal (NORM) groups
Full size table

4 Discussion

The purpose of this study was to determine how a concussion affects inter-joint coordination during level walking with and without a concurrent cognitive task. Our results showed that the kinematic control patterns of each individual joint might be altered following a concussion, especially at the knee and ankle joints. Patterns of the inter-joint coordination between the hip and the knee and between the knee and the ankle during walking with or without a concurrent cognitive task for concussed subjects were similar to those of normal individuals. However, a greater variability in swing phase hip–knee coordination was observed in individuals who had had a concussion.

All subjects demonstrated a slower gait velocity and a shorter stride length during dual-task walking. The simultaneous execution of a mental task increased the challenge of the walking task since motor and mental tasks compete for limited attention capacity [30]. It was found that in order to achieve performance similar to that of normal individuals in cognitive tasks, individuals who had suffered a concussion had to reduce their gait velocities at a greater rate than that of their matched controls (12.8 vs. 8.8 %). One possible explanation is that concussion decreases the attention capacity of an individual, resulting in altered motor performance while maintaining similar performance on the concurrent cognitive task during walking. However, it is also likely that the ability to allocate attention between concurrent tasks is compromised by concussion.

Apart from changes in temporal–spatial variables, concussion also affected the joint kinematics of the lower extremities during both task conditions. The joint phase angle indicates the coordination of the segments connected by the joint. A comparison of the phase angle plots indicates that the greatest CMC values were found at the hip joint, which could indicate that the concussion group was able to maintain a hip motion pattern similar to that of the control group. However, the level of similarity was found to decrease at the knee and ankle joints. Only the ankle joint is directly involved in controlling foot motion and might thus contribute more to the changes in gait velocity. This finding suggests that individuals who had had a concussion might exhibit a distal adaptation strategy, in terms of altered knee and ankle joint motion patterns, to achieve the goal of level walking with and without divided attention tasks.

Individuals who had had a concussion altered distal joint kinematic control during walking with and without divided attention tasks. However, this does not change the way that two adjacent joints are coordinated. The CMC values of the CRP curves between CONC and NORM groups were all greater than 0.8, suggesting that the inter-joint coordination patterns between the two groups were similar. This finding, combined with findings from previous studies [18, 19], indicates that changes in joint motion or gait performance in individuals with brain or joint dysfunction are regulated by the central nervous system to maintain coordination patterns between joints. To maintain consistent inter-joint coordination patterns, changes in gait temporal–spatial variables as well as the motion pattern of more distal joints may be sufficient to compromise for the effects of mild brain injury.

Although the inter-joint coordination pattern did not seem to be affected by concussion, its variability was increased. As described by Kelso et al. [31], the human body can be modeled as a complex neuro-controller coupled with a non-linear dynamics system. Any variability of the movement trajectories generated by this system is a product of input disturbances, neuromuscular control, and biomechanical dynamics. Individuals who had had a concussion may have a reduced attention capacity, possibly leading to difficulties in executing a single task or easily switching their attention between two simultaneous tasks. This may increase the central processing demand of the brain and thus induce more variability in movement control. The DP value of the swing phase hip–knee was greater in concussed subjects, which indicates that they need greater control effort to coordinate the swing phase hip–knee motion. This result could explain the findings of Mcfadyen et al. [32], who found that TBI affected the intra-subject variability in clearance margins. Increased variability of toe clearance following TBI possibly results from the increased variability of hip–knee coordination. It is interesting that changes in the knee and ankle joint motion patterns did not affect the knee–ankle inter-joint coordination pattern and variability. The changes in the motion of the knee and the ankle in the CONC group may be due to not only the adaptation of gait temporal–spatial changes (i.e., slower gait velocity) but also the maintenance of consistent coordination between knee and ankle joints. In contrast, the greater DP value of the swing phase hip–knee with decreased CMC value of the knee phase angle suggests that to maintain a consistent coordination pattern with hip motion similar to NORM group requires greater effort in CONC group.

5 Conclusion

The effect of concussion on the inter-joint coordination of the lower extremities during walking was investigated. It is possible that if injury to the central nervous system does not occur, such as with knee osteoarthritis, neither the pattern nor variability of inter-joint coordination would be affected. However, if the injury or degeneration is associated with the central nervous system, such as with aging or concussion, the inter-joint coordination variability is affected but not the coordination patterns. Therefore, examination of inter-joint coordination may provide insight into the observed deviations in gait performance and joint motion of individuals who had suffered a concussion. Exercises for improving inter-joint coordination, especially hip–knee coordination, may help prevent these patients from recurrent concussions. For athletes who participate in contact sports, evaluation of inter-joint coordination may serve as a criterion for the safe return to play following a concussion.