An Artificial Neural Network Estimation of Gait Balance Control in the Elderly Using Clinical Evaluations

The use of motion analysis to assess balance is essential for determining the underlying mechanisms of falls during dynamic activities. Clinicians evaluate patients using clinical examinations of static balance control, gait performance, cognition, and neuromuscular ability. Mapping these data to measures of dynamic balance control, and the subsequent categorization and identification of community dwelling elderly fallers at risk of falls in a quick and inexpensive manner is needed. The purpose of this study was to demonstrate that given clinical measures, an artificial neural network (ANN) could determine dynamic balance control, as defined by the interaction of the center of mass (CoM) with the base of support (BoS), during gait. Fifty-six elderly adults were included in this study. Using a feed-forward neural network with back propagation, combinations of five functional domains, the number of hidden layers and error goals were evaluated to determine the best parameters to assess dynamic balance control. Functional domain input parameters included subject characteristics, clinical examinations, cognitive performance, muscle strength, and clinical balance performance. The use of these functional domains demonstrated the ability to quickly converge to a solution, with the network learning the mapping within 5 epochs, when using up to 30 hidden nodes and an error goal of 0.001. The ability to correctly identify the interaction of the CoM with BoS demonstrated correlation values up to 0.89 (P<.001). On average, using all clinical measures, the ANN was able to estimate the dynamic CoM to BoS distance to within 1 cm and BoS area to within 75 cm². Our results demonstrated that an ANN could be trained to map clinical variables to biomechanical measures of gait balance control. A neural network could provide physicians and patients with a cost effective means to identify dynamic balance issues and possible risk of falls from routinely collected clinical examinations.     

Age-related reduction in sagittal plane center of mass motion during obstacle crossing

Accidental falls are a leading cause of injury and death in the growing elderly population. Traumatic falls are frequent, costly, and debilitating. Control of balance during locomotion is critical for safe ambulation, but relatively little is known about the natural effect of aging on dynamic balance control. Samples of healthy young (n = 13) and elderly (n = 13) subjects were compared in the interactive measures of center of mass (COM) and center of pressure (COP) during level walking and obstacle crossing conditions. Obstacle heights were normalized to individual body height (2.5%, 5%, 10%, and 15%). Temporal-distance (T-D) variables of gait were also compared. Statistical analyses were conducted using a two-way ANOVA for subject group and obstacle height. T-D parameters were not significantly different between groups; nor were frontal plane COM and COP parameters. Significant age differences did exist for antero-posterior (A/P) motion of the COM (decreased motion in the elderly), and its relationship with the COP (reduced separation between the two variables in the elderly). Anterior COM velocities were also significantly lower in the elderly group. The results confirm the ability of healthy elderly adults to maintain dynamic balance control in the frontal plane during locomotion. Reduced A/P distances between the COM and COP indicate a conservative reduction of the mechanical load on joints of the supporting limb. This conservative strategy may be related to a reduction in muscle strength as it occurs in the natural aging process.     

Feasibility of estimating isokinetic knee torque using a neural network model

Many studies have investigated the relationships between electromyography (EMG) and torque production. A few investigators have used adjusted learning algorithms and feed-forward artificial neural networks (ANNs) to estimate joint torque in the elbow. This study sought to estimate net isokinetic knee torque using ANN models. Isokinetic knee extensor and flexor torque data were measured simultaneously with agonist and antagonist EMG during concentric and eccentric contractions at joint velocities of 30 degrees /s and 60 degrees /s. Age, gender, height, body mass, agonist EMG, antagonist EMG, joint position and joint velocity were entered as predictive variables of net torque. A three-layer ANN model was developed and trained using an adjusted back-propagation algorithm. Accuracy results were compared against those of forward stepwise regression models. Stepwise regression models included body mass, body height and joint position as the most influential predictors, followed by agonist EMG for concentric and eccentric contractions. Estimation of eccentric torque included antagonist EMG following the agonist activation. ANN models resulted in more accurate torque estimation (R=0.96), compared to the stepwise regression models (R=0.71). ANN model accuracy increased greatly when the number of hidden units increased from 5 to 10, continuing to increase gradually with additional hidden units. The average number of training epochs necessary for solution convergence and the relative accuracy of the model indicate a strong ability for the ANN model to generalize these estimations to a broader sample. The ANN model appears to be a feasible technique for estimating joint torque in the knee.     

Does the anthropometric model influence whole-body center of mass calculations in gait?

Examining whole-body center of mass (COM) motion is one of method being used to quantify dynamic balance and energy during gait. One common method for estimating the COM position is to apply an anthropometric model to a marker set and calculate the weighted sum from known segmental COM positions. Several anthropometric models are available to perform such a calculation. However, to date there has been no study of how the anthropometric model affects whole-body COM calculations during gait. This information is pertinent to researchers because the choice of anthropometric model may influence gait research findings and currently the trend is to consistently use a single model. In this study we analyzed a single stride of gait data from 103 young adult participants. We compared the whole-body COM motion calculated from 4 different anthropometric models (Plagenhoef et al., 1983; Winter, 1990; de Leva, 1996; Pavol et al., 2002). We found that anterior-posterior motion calculations are relatively unaffected by the anthropometric model. However, medial-lateral and vertical motions are significantly affected by the use of different anthropometric models. Our findings suggest that the researcher carefully choose an anthropometric model to fit their study populations when interested in medial-lateral or vertical motions of the COM. Our data can provide researchers a priori information on the model determination depending on the particular variable and how conservative they may want to be with COM comparisons between groups.     

The effect of mood phases on balance control in bipolar disorder

The aim of this study was to investigate balance control during gait and sit-to-walk in individuals with bipolar disorder and healthy controls by examining the inclination angles between the whole-body center-of-mass (COM) and ankle in the sagittal plane. Twenty-one individuals with bipolar disorder in the euthymic (i.e., asymptomatic; n = 11) and depressed (n = 10) phases and 7 healthy controls (ages between 18 and 45) performed gait and sit-to-walk at self-selected comfortable speed. Mood phases for individuals with bipolar disorder were measured using the Patient Health Questionnaire and Altman Self-Rating Mania Scale. We collected motion data using a 16-camera motion capture technology. We found smaller COM-ankle inclination angles at seat-off during sit-to-walk for the bipolar-depressed group compared to the bipolar-euthymic and healthy groups, indicating poorly controlled balance for the bipolar-depressed group in sit-to-walk. However, we found larger COM-ankle inclination angles at beginning of single stance phase of gait for the bipolar-euthymic group compared to the healthy group, indicating well controlled balance for the bipolar-euthymic group in gait. Our results suggest an association between the depressed phase and balance impairment during daily movements in relatively young adults (ages ≤ 45 years). Our results also suggest that the depressed phase may be as detrimental to balance control as the effect of age-related neuromuscular weakness.     

Responsiveness to rehabilitation of balance and gait impairment in elderly with peripheral neuropathy

Elderly people with peripheral neuropathy of the lower limbs (PNLL) demonstrate a typical balance and gait impairment because of sensory ataxia. There is evidence that rehabilitation produces important gains on balance and gait. However, responsiveness to rehabilitation of balance and gait measures is unknown in PNLL. Aim of the current work is to evaluate the responsiveness to rehabilitation of balance, gait and sensory ataxia measures in elderly with PNLL.Twenty-five elderly with PNLL attending physiotherapy and occupational therapy during inpatient rehabilitation were recruited. Balance and gait measures (including static posturography, TUG test and the 10 m walking test) were administered on admission and discharge. An accelerometer secured to the trunk was used for TUG recording and static balance assessment. Static balance was tested with open and closed eyes, so as to assess sensory ataxia.Following rehabilitation, patients improved gait [admission vs discharge, mean(SD): 0.86(0.33) vs 0.98(0.32) m/s], TUG [18.7(7.8) vs 15.1(5.2) s] and turning [46.2(15.3) vs 53.3(15.3) °/s]. However, none of 12 static balance parameters derived from trunk acceleration significantly changed. Principal component analysis showed that before training, eyes closed and eyes open balance correlated with orthogonal components (one and two vs. three and four). After training, eyes open and eyes closed balance were more similar to each other being both correlated with component one.Responsiveness to rehabilitation is larger for gait than static balance measured by trunk acceleration. However, exercise can also have a beneficial effect on sensory ataxia by making eyes closed balance more similar to eyes open balance.     

Dynamic balance control during sit-to-stand movement: an examination with the center of mass acceleration

The purpose of this study was to establish the region of stability of balance control using the center of mass (COM) acceleration and to characterize age-related differences during sit-to-stand (STS) movement. Whole body motion data were collected from 10 young and 10 elderly subjects while performing STS at their self-selected manners. In addition, young subjects were asked to perform another block of trials with their trunk purposely bent forward prior to seat-off. With the use of a single-link-plus-foot inverted pendulum model, boundaries for the region of stability were determined based on the COM position at seat-off and its instantaneous velocity or its peak acceleration (ROSv or ROSa, respectively). No significant group differences were detected in COM velocities at seat-off. However, peak COM accelerations differed significantly between groups and conditions. This suggested that even though a similar COM momentum was observed at seat-off, this momentum was controlled differently prior to seat-off. Young and elderly subjects utilized similar strategies but with different COM acceleration profiles to perform STS. Furthermore, data from an elderly subject who complained of difficulty in STS during the experiment were located outside the forward boundary of the ROSa, demonstrating a potential use of ROSa to differentiate individuals with declined balance control ability. The ROSa could provide insights into how the COM is controlled prior to seat-off, which may allow us to better identify elderly individuals who are most likely at a risk for imbalance or falls.     

Age-related changes in posture response under a continuous and unexpected perturbation

Aging is a critical factor to influence the functional performance during daily life. Without an appropriate posture control response when experiencing an unexpected external perturbation, fall may occur. A novel six-degree-of freedom platform with motion control protocol was designed to provide a real-life simulation of unexpected disturbance in order to discriminate the age-related changes of the balance control and the recovery ability. Twenty older adults and 20 healthy young adults participated in the study. The subjects stood barefoot on the novel movable platform, data of the center of mass (COM) excursion, joint rotation angle and electromyography (EMG) were recorded and compared. The results showed that the older adults had similar patterns of joint movement and COM excursion as the young adults during the balance reactive-recovery. However, larger proximal joint rotation in elderly group induced larger COM sway envelop and therefore loss of the compensatory strategy of posture recovery. The old adults also presented a lower muscle power. In order to keep an adequate joint stability preventing from falling, the EMG activity was increased, but the asymmetric pattern might be the key reason of unstable postural response. This novel design of moveable platform and test protocol comprised the computerized dynamic posturography (CDP) demonstrate its value to assess the possible sensory, motor, and central adaptive impairments to balance control and could be the training tool for posture inability person.     

Predicted threshold against backward balance loss following a slip in gait

The purpose of this study was to use a 7-link, moment-actuated human model to predict, at liftoff of the trailing foot in gait, the threshold of the center of mass (COM) velocity relative to the base of support (BOS) required to prevent backward balance loss during single stance recovery from a slip. Five dynamic optimization problems were solved to find the minimum COM velocities that would allow the simulation to terminate with the COM above the BOS when the COM started 0.25, 0.5, 0.75, 1.0, and 1.25 foot lengths behind the heel of the stance foot (i.e., behind the BOS). The initial joint angles of the model were based on averaged data from experimental trials. Foot-ground contact was modeled using 16 visco-elastic springs distributed under the stance foot. Slipping was modeled by setting the sliding coefficient of friction of these springs to 0.02. The forward velocity of the COM necessary to avoid a backward balance loss is nearly two times larger under slip conditions under non-slip conditions. The predicted threshold for backward balance loss following a slip agreed well with experimental data collected from 99 young adults in response to 927 slips during walking. In all trials in which a subject's COM had a velocity below the predicted threshold, the subject's recovery foot landed posterior to the slipping foot as predicted. Finally, combining experimental data with optimization, we verified that the 7-link model could more accurately predict gait stability than a 2-link model.     

Predicted threshold against backward balance loss in gait

The purpose of this study was to determine the minimum forward center of mass (COM) velocity required to prevent backward loss of balance in gait as function of the initial COM position. We hypothesized that these threshold values would be different from those previously published for standing because of the postural differences between gait and standing. To investigate this issue, we constructed a seven-link, nine-degree-of-freedom biomechanical model and employed dynamic optimization to estimate these threshold values under two initial postural conditions: (1) the posture at the beginning of swing phase (i.e., at toe-off), and (2) symmetric bipedal standing. In particular, for a range of possible COM positions posterior to the base of support (BOS), simulated annealing was used to search for the minimum velocity that could carry the COM into the BOS and avoid backward loss of balance. We found that the stability boundary against backward balance loss in walking had a similar overall trend as that previously published for standing. In general, the minimal COM velocity necessary to prevent a backward loss of balance in walking was greater than that in symmetric bipedal standing, and the difference could approach 30% or more when the COM started 0.5 and 1.0 foot-lengths behind the BOS. These discrepancies suggest that simpler biomechanical models, while being more efficient and easier to employ, may not always be adequate for exploring stability limits of humans.     

Can sacral marker approximate center of mass during gait and slip-fall recovery among community-dwelling older adults?

Falls are prevalent in older adults. Dynamic stability of body center of mass (COM) is critical for maintaining balance. A simple yet accurate tool to evaluate COM kinematics is essential to examine the COM stability. The purpose of this study was to determine the extent to which the COM position derived from body segmental analysis can be approximated by a single (sacral) marker during unperturbed (regular walking) and perturbed (gait-slip) gait. One hundred eighty seven older adults experienced an unexpected slip after approximately 10 regular walking trials. Two trials, the slip trial and the preceding regular walking trial, monitored with a motion capture system and force plates, were included in the present study. The COM positions were calculated by using the segmental analysis method wherein, the COM of all body segments was calculated to further estimate the body COM position. These body COM positions were then compared with those of the sacral marker placed at the second sacral vertebra for both trials. Results revealed that the COM positions were highly correlated with those of the sacrum׳s over the time intervals investigated for both walking (coefficient of correlation R>0.97) and slip (R>0.90) trials. There were detectable kinematic difference between the COM and the sacral for both trials. Our results indicated that the sacral marker can be used as a simple approximation of body COM for regular walking, and to somewhat a lesser extent, upon a slip. The benefits from the simplicity appear to overweigh the limitations in accuracy.     

Adaptive control of center of mass (global) motion and its joint (local) origin in gait

Dynamic gait stability can be quantified by the relationship of the motion state (i.e. the position and velocity) between the body center of mass (COM) and its base of support (BOS). Humans learn how to adaptively control stability by regulating the absolute COM motion state (i.e. its position and velocity) and/or by controlling the BOS (through stepping) in a predictable manner, or by doing both simultaneously following an external perturbation that disrupts their regular relationship. Post repeated-slip perturbation training, for instance, older adults learned to forward shift their COM position while walking with a reduced step length, hence reduced their likelihood of slip-induced falls. How and to what extent each individual joint influences such adaptive alterations is mostly unknown. A three-dimensional individualized human kinematic model was established. Based on the human model, sensitivity analysis was used to systematically quantify the influence of each lower limb joint on the COM position relative to the BOS and the step length during gait. It was found that the leading foot had the greatest effect on regulating the COM position relative to the BOS; and both hips bear the most influence on the step length. These findings could guide cost-effective but efficient fall-reduction training paradigm among older population.     

Control of center of mass motion state through cuing and decoupling of spontaneous gait parameters in level walking

Can the center of mass (COM) motion state, i.e., its position and velocity relative to the base of support (BOS), which dictate gait stability, be predictably controlled by the global gait parameters of step length and gait speed, or by extension, cadence? The precise relationships among step length and gait speed, and the COM motion state are unknown, partially due to the interdependence between step length and gait speed and the difficulty in independent control of both parameters during spontaneous level walking. The purposes of this study were to utilize simultaneous audio-visual cuing to independently manipulate step length and gait speed, and to determine the extent to which the COM position and velocity can be subsequently controlled. Fifty-six young adults were trained at one of the three gait patterns in which both the step length and gait speed were targeted simultaneously. The results showed that the cuing could successfully “decouple” gait speed from step length. Although this approach did yield reliable control of the COM velocity through manipulation of gait speed (R²=0.97), the manipulation of step length yielded less precise control of COM position (R²=0.60). This latter control appears to require manipulation of an additional degree-of-freedom at the local segment level, such that the inclusion of trunk inclination with step length improved the prediction of COM position (R²=0.80).     

The effect of a 12-week combined exercise intervention program on physical performance and gait kinematics in community-dwelling elderly women

This study aimed to determine if combined exercise intervention improves physical performance and gait joint-kinematics including the joint angle and dynamic range of motion (ROM) related to the risk of falling in community-dwelling elderly women. A 12-week combined exercise intervention program with extra emphasis on balance, muscle strength, and walking ability was designed to improve physical performance and gait. Twenty participants attended approximately two-hour exercise sessions twice weekly for 12 weeks. Participants underwent a physical performance battery, including static balance, sit and reach, whole body reaction time, 10 m obstacle walk, 10 m maximal walk, 30-second chair stand, to determine a physical performance score, and received quantitative gait kinematics measurements at baseline and in 12 weeks. Significant lower extremity strength improvement 13.5% (p<.001) was observed, which was accompanied by significant decreases in time of the 10 m obstacle walk (p<.05) and whole body reaction time (p<.001) in this study. However, no significant differences were seen for static balance and flexibility from baseline. For gait kinematics, in the mid-swing phase, knee and hip joint angle changed toward flexion (p<.01, p<.05, respectively). Ankle dynamic ROM significantly increased (p<.05) following exercise intervention. The plantar flexion angle of the ankle in the toe-off phase was increased significantly (p<.01). However, other gait parameters were not significantly different from baseline. These findings from the present investigation provide evidence of significant improvements in physical performance related to the risk factors of falling and safe gait strategy with a combined exercise intervention program in community-dwelling elderly women. The results suggest this exercise intervention could be an effective approach to ameliorate the risk factors for falls and to promote safer locomotion in elderly community-dwelling women.     

Inertial motion capture in conjunction with an artificial neural network can differentiate the gait patterns of hemiparetic stroke patients compared with able-bodied counterparts

Clinical gait analysis has proven to reduce uncertainties in selecting the appropriate quantity and type of treatment for patients with neuromuscular disorders. However, gait analysis as a clinical tool is under-utilised due to the limitations and cost of acquiring and managing data. To overcome these obstacles, inertial motion capture (IMC) recently emerged to counter the limitations attributed to other methods. This paper investigates the use of IMC for training and testing a back-propagation artificial neural network (ANN) for the purpose of distinguishing between hemiparetic stroke and able-bodied ambulation. Routine gait analysis was performed on 30 able-bodied control subjects and 28 hemiparetic stroke patients using an IMC system. An ANN was optimised to classify the two groups, achieving a repeatable network accuracy of 99.4%. It is concluded that an IMC system and appropriate computer methods may be useful for the planning and monitoring of gait rehabilitation therapy of stroke victims.     

Inertial motion capture in conjunction with an artificial neural network can differentiate the gait patterns of hemiparetic stroke patients compared with able-bodied counterparts

Clinical gait analysis has proven to reduce uncertainties in selecting the appropriate quantity and type of treatment for patients with neuromuscular disorders. However, gait analysis as a clinical tool is under-utilised due to the limitations and cost of acquiring and managing data. To overcome these obstacles, inertial motion capture (IMC) recently emerged to counter the limitations attributed to other methods. This paper investigates the use of IMC for training and testing a back-propagation artificial neural network (ANN) for the purpose of distinguishing between hemiparetic stroke and able-bodied ambulation. Routine gait analysis was performed on 30 able-bodied control subjects and 28 hemiparetic stroke patients using an IMC system. An ANN was optimised to classify the two groups, achieving a repeatable network accuracy of 99.4%. It is concluded that an IMC system and appropriate computer methods may be useful for the planning and monitoring of gait rehabilitation therapy of stroke victims.     

Biomechanical mechanism of lateral trunk lean gait for knee osteoarthritis patients

The biomechanical mechanism of lateral trunk lean gait employed to reduce external knee adduction moment (KAM) for knee osteoarthritis (OA) patients is not well known. This mechanism may relate to the center of mass (COM) motion. Moreover, lateral trunk lean gait may affect motor control of the COM displacement. Uncontrolled manifold (UCM) analysis is an evaluation index used to understand motor control and variability of the motor task. Here we aimed to clarify the biomechanical mechanism to reduce KAM during lateral trunk lean gait and how motor variability controls the COM displacement. Twenty knee OA patients walked under two conditions: normal and lateral trunk lean gait conditions. UCM analysis was performed with respect to the COM displacement in the frontal plane. We also determined how the variability is structured with regards to the COM displacement as a performance variable. The peak KAM under lateral trunk lean gait was lower than that under normal gait. The reduced peak KAM observed was accompanied by medially shifted knee joint center, shortened distance of the center of pressure to knee joint center, and shortened distance of the knee–ground reaction force lever arm during the stance phase. Knee OA patients with lateral trunk lean gait could maintain kinematic synergy by utilizing greater segmental configuration variance to the performance variable. However, the COM displacement variability of lateral trunk lean gait was larger than that of normal gait. Our findings may provide clinical insights to effectively evaluate and prescribe gait modification training for knee OA patients.     

A machine learning approach for automated recognition of movement patterns using basic, kinetic and kinematic gait data

This paper investigated application of a machine learning approach (Support vector machine, SVM) for the automatic recognition of gait changes due to ageing using three types of gait measures: basic temporal/spatial, kinetic and kinematic. The gaits of 12 young and 12 elderly participants were recorded and analysed using a synchronized PEAK motion analysis system and a force platform during normal walking. Altogether, 24 gait features describing the three types of gait characteristics were extracted for developing gait recognition models and later testing of generalization performance. Test results indicated an overall accuracy of 91.7% by the SVM in its capacity to distinguish the two gait patterns. The classification ability of the SVM was found to be unaffected across six kernel functions (linear, polynomial, radial basis, exponential radial basis, multi-layer perceptron and spline). Gait recognition rate improved when features were selected from different gait data type. A feature selection algorithm demonstrated that as little as three gait features, one selected from each data type, could effectively distinguish the age groups with 100% accuracy. These results demonstrate considerable potential in applying SVMs in gait classification for many applications.     

Balance control during stair negotiation in older adults

Stair negotiation is among the most challenging and hazardous types of locomotion for older people. However, the effect of aging on balance control during stair negotiation has not been investigated. Instantaneous inclination angles between the center of mass (CoM) and center of pressure (CoP) have been reported to detect gait instability effectively in the elderly. The purpose of this study was to compare the CoM-CoP inclination angles between 12 healthy elderly and 13 healthy young adults when performing stair ascent (SA) and descent (SD) on a three-step staircase. Whole body motion data were collected with an eight-camera motion analysis system. Four force plates were mounted on the floor as well as the first two steps to measure ground reaction forces. No significant group differences were detected in any of the temporal-distance gait measures and CoM-CoP inclination angles during SA and SD. Compared to the floor-to-stair transition phase, both groups demonstrated a significantly greater CoM-CoP medial inclination angle while ascending the stairs. However, a significant reduction in medial inclination was only detected in young adults when transferring from SD to level ground walking. Elderly adults were found to demonstrate a significantly greater medial inclination angle during the stair-to-floor transition phase when compared to young adults. Age-related degenerations in the elderly could compromise their ability to regulate body sway during the stair-to-floor transition, which may subsequently increase the risk of falling.     

Statistical method for prediction of gait kinematics with Gaussian process regression

We propose a novel methodology for predicting human gait pattern kinematics based on a statistical and stochastic approach using a method called Gaussian process regression (GPR). We selected 14 body parameters that significantly affect the gait pattern and 14 joint motions that represent gait kinematics. The body parameter and gait kinematics data were recorded from 113 subjects by anthropometric measurements and a motion capture system. We generated a regression model with GPR for gait pattern prediction and built a stochastic function mapping from body parameters to gait kinematics based on the database and GPR, and validated the model with a cross validation method. The function can not only produce trajectories for the joint motions associated with gait kinematics, but can also estimate the associated uncertainties. Our approach results in a novel, low-cost and subject-specific method for predicting gait kinematics with only the subject's body parameters as the necessary input, and also enables a comprehensive understanding of the correlation and uncertainty between body parameters and gait kinematics.