Reactive Muscle Force Modulation Following Induced Belt Deceleration in Treadmill Gait

INTRODUCTION

Gait is considered the most fundamental and natural form of human locomotion [1]. Despite its apparent simplicity, it constitutes one of the most complex motor activities. The act of walking involves the repetitive and finely controlled loss and recovery of balance during alternating phases of support and swing, while simultaneously advancing the body’s center of gravity in the forward direction. This dynamic process is mechanically demanding, as in the upright position, the human body behaves similarly to an inverted pendulum, requiring continuous adjustments to maintain equilibrium [2,3]. Consequently, maintaining stability in this posture poses a considerable challenge, particularly in the presence of external perturbations.

During walking, two primary forms of perturbations are typically distinguished: slips and trips [4]. A slip occurs when the frictional force required for forward progression exceeds the available friction between the walking surface and the foot [5]. Conversely, a trip takes place when the swing phase of the foot is disrupted due to insufficient ground clearance [6], often leading to a loss of balance and, in many cases, falls [7].

To counteract these perturbations, the human neuromuscular system engages a range of rapid and coordinated compensatory strategies [8]. Recovery from disturbance is achieved through adjustments in joint kinematics and kinetics, repositioning of the trunk, and corrective stepping responses collectively referred to as recovery steps [9]. The selection and execution of an appropriate strategy are governed by an intricate interaction between spinal and brainstem reflexes, task-dependent sensorimotor integration, and supraspinal control mechanisms that together regulate the timing, amplitude, and coordination of muscle responses [10]. The magnitude and direction of the perturbation further influence which compensatory strategy is adopted and how it is executed.

Reactions to gait perturbations have been extensively examined in previous research [11]. Beyond the analysis of kinematic, kinetic, and spatiotemporal parameters, particular attention has been devoted to muscle responses during recovery. Moreover, differences in reaction time, a key element in balance control and motor response initiation, were observed in populations with varying levels of physical activity, highlighting the importance of neuromotor abilities in adapting to sudden disturbances [12]. Beyond the analysis of kinematic, kinetic, and spatiotemporal parameters, particular attention has been devoted to muscle responses during recovery. Electromyography (EMG) has been the predominant technique used to explore the activation patterns of specific muscles [13]. However, EMG-based analyses have several notable limitations. First, only a limited number of muscles can be assessed simultaneously, restricting a comprehensive understanding of the muscle-specific coordination patterns involved in perturbation recovery. Second, EMG provides information about neural activation but not about the resulting muscle forces [14]. This distinction is critical, as the ability to generate adequate muscle force, rather than activation alone, ultimately determines the success of balance recovery.

A considerable portion of prior research has focused on older adults, given their increased susceptibility to falls [15]. Nonetheless, the examination of perturbation responses in young, healthy individuals remains essential. Studying this baseline population, free from age-related sensory decline, muscular weakness, or neurological impairment, enables a clearer understanding of the fundamental neuromuscular and biomechanical mechanisms underlying reactive balance control. In biomechanical analyses of responses to movement disorders, it is also important to consider limb dominance, as it affects postural symmetry and balance control, as demonstrated in recent studies of young adult populations [16]. Furthermore, defining normative characteristics of these mechanisms in young adults provides a valuable reference for comparison with older or clinical populations. Such knowledge also contributes to the development of evidence-based interventions, including perturbation-based balance training programs, aimed at improving stability and preventing falls [17,18]. Analysis of muscle forces, particularly peak muscle forces, offers valuable insight into the mechanical demands imposed on individual muscles during reactive balance recovery. Unlike EMG, which reflects only the neural drive to the muscles, muscle force estimation integrates both neural and mechanical factors, such as muscle length, contraction velocity, and moment arm geometry, that together determine the actual output of the muscle-tendon unit [19-21]. Accordingly, the assessment of peak muscle forces allows for a more direct evaluation of the contributions of specific muscles to the compensatory response following perturbation [22]. Identifying which muscles experience the highest loads during recovery can further inform the design of targeted prevention and training strategies aimed at enhancing reactive stability and minimizing fall risk [23].

To date, studies directly examining the behavior of individual muscle forces during gait perturbation have been lacking [24]. Therefore, the present study aimed to investigate changes in peak lower-limb muscle forces during the recovery step following a perturbation, compared with those observed during unperturbed walking, in healthy young adults.

MATERIAL AND METHODS

Trial and Perturbation Protocol

Kinematic (joint angles) and kinetic (joint torques and ground reaction forces) data for both perturbed and unperturbed gait were collected in the Interactive Gait Real-Time Analysis Laboratory (GRAIL, Motek Medical B.V., Amsterdam, The Netherlands). The GRAIL system included a dual-belt treadmill (2 × 500 mm width, 2200 mm length) operating at 1000 Hz, a ten-camera Vicon motion capture system (Bonita cameras; Vicon Metrics Ltd., Oxford, UK) recording at 100 Hz, three synchronized video cameras, and a 180° virtual reality projection screen (5 m diameter, 3 m height) supported by three projectors. All data were acquired and synchronized using D-Flow v. 3.26 software (Motek Medical B.V., Amsterdam, The Netherlands), which was also employed for model calibration and the control of perturbations.

Upon arrival, participants’ anthropometric measurements were recorded, and twenty-five reflective markers were placed on anatomical landmarks according to the Human Body Model 2 (HBM2) protocol (Figure 1). Participants wore their own comfortable athletic footwear, typically used for treadmill walking or gym activities. They were then positioned on the treadmill in a standardized T-pose to ensure precise alignment of the digital body model within D-Flow. A ceiling-mounted safety harness was used to secure participants and prevent falls throughout the trials.

Each participant completed a walking trial lasting 80 seconds, during which the treadmill speed was maintained at a constant 1.2 m/s. Within the trial, five perturbations were delivered at the 30th, 40th, 50th, 60th, and 70th seconds. Perturbations were set to level 5 on a scale of 1 to 5, corresponding to a sudden decrease in treadmill belt velocity of approximately 0.5 – 0.6 m/s. As a result, the speed of the left treadmill belt dropped to 0.6

– 0.7 m/s during perturbation events (Figure 2).

The perturbation protocol followed the methodology described by Sloot, et al. [28] and Chodkowska, et al. [24], with each perturbation lasting approximately 0.93 s. Perturbations were applied exclusively to the left treadmill belt, targeting the non-dominant (left) lower limb. The highest perturbation magnitude was selected to ensure a distinct and measurable response, allowing accurate estimation of peak joint and muscle loads. Perturbations were delivered during the Pre-Swing (PSw) phase of the gait cycle, corresponding to approximately 50 – 60% of the cycle as defined by Perry and Burnfield [29]. The D-Flow software enabled precise control of perturbation parameters, including magnitude, treadmill lane, timing, and event type. Since gait phases represent intervals rather than precise moments in time, a perturbation applied, for instance, at the 30th second during the PSw phase may correspond to slightly different points within the gait cycle across participants.

Workflow for Musculoskeletal Simulation in OpenSim

In the first phase of data processing, motion data exported from D-Flow v3.26 were initially saved as *.mox files and imported into MATLAB 2021a (MathWorks, Natick, MA, USA) using a toolbox developed by Feldhege, et al. [30]. The imported data were then converted into OpenSim-compatible input files: *.trc files, containing three-dimensional marker trajectories captured during motion recording, and *.mot files, containing synchronized ground reaction force (GRF) and torque data [19]. Conversion scripts are available in public repository: https://github.com/RafalSci/Muscle-force-modulation-after-gait-perturbations.git.

For simulating muscle force contributions during both unperturbed walking and responses to perturbations, the Gait2392 musculoskeletal model (Fig. 3A), available in the OpenSim library, was used. This model is a three-dimensional, 23-degree-of-freedom representation of the human body comprising 92 musculotendon actuators, which represent 30 muscles controlling a single lower limb and 3 muscles acting on the torso on one side. The generic musculoskeletal model was then scaled to match each participant’s anthropometry using the OpenSim Scale Tool. Scaling was performed based on static calibration trials to adjust segment lengths and marker placements, ensuring accurate correspondence with each participant’s body geometry.

Following model scaling, inverse kinematics (IK) analysis was conducted to calculate joint angles across the trial. The IK algorithm aligns the model’s virtual markers with experimentally recorded marker positions at each time frame, minimizing the sum of squared positional errors through a least-squares optimization approach. This process yields precise joint angle trajectories that describe three-dimensional limb and pelvis motion during both perturbed and unperturbed walking.

Inverse dynamics (ID) analysis was then performed to compute joint torques and joint reaction forces. This step applies Newton–Euler equations of motion to determine the net torques acting at each joint, based on the derived kinematic data, segment inertial properties, and external GRFs. The resulting joint torques provide insight into the mechanical demands imposed on the musculoskeletal system during normal gait and perturbation recovery, reflecting the total effort required to counteract internal and external forces.

In the final stage, static optimization (SO) was used to estimate individual muscle forces at each time frame. This algorithm distributes the net joint moments obtained from inverse dynamics among the model’s muscles by minimizing the sum of squared muscle activations while satisfying physiological and mechanical constraints. The SO approach assumes quasi-static equilibrium at each instant of motion, allowing efficient computation of physiologically realistic muscle force patterns. Through this method, it was possible to identify which muscles contribute most significantly to maintaining stability and generating propulsion during perturbed gait (Fig. 3B).

RESULTS

Across all examined muscles, peak muscle force (PMF) values recorded during the recovery step were higher compared to those observed during unperturbed walking. The magnitude of these increases varied across muscles, ranging from 7.26% in the m. Peroneus brevis to 118.86% in the m. Tensor fasciae latae (Table 1).

The k-means cluster analysis (n = 3) grouped muscles according to the magnitude of ΔPMF, revealing three distinct activation patterns. Cluster 1, representing high response (ΔPMF ≥ 100%), consisted of three muscles primarily responsible for proximal hip stabilization, with the largest increase observed in the Tensor fasciae latae (118.86%, p < 0.001) and the smallest in the Gluteus maximus (101.37%, p < 0.001). Cluster 2, representing moderate response (ΔPMF = 40 – 100%), included ten muscles contributing to hip and knee stabilization and lower-limb support, where the highest increase occurred in the Tibialis posterior (78.39%, p < 0.001) and the lowest in the Gluteus minimus (40.56%, p < 0.001). Cluster 3, representing low response (ΔPMF ≤ 40%), comprised 18 muscles, mainly distal or postural muscles, which exhibited smaller adjustments, with the highest increase seen in the Extensor digitorum (32.70%, p < 0.05) and the lowest in the Peroneus brevis (7.26%, not significant).

These findings indicate that recovery from a perturbation relies predominantly on proximal stabilizers and key lower-limb muscles, which generate the largest increases in force to rapidly restore balance. In contrast, distal and postural muscles show smaller adjustments, suggesting a hierarchical, coordinated neuromuscular strategy in which proximal muscles act as primary stabilizers while distal muscles provide fine-tuned support for postural control.

DISCUSSION

The present study investigated changes in peak muscle forces (PMF) during the recovery step in response to unexpected treadmill belt decelerations applied in the pre-swing (PSw) phase of the gait cycle in healthy young females using biomechanical modeling with OpenSim software. The main finding was a substantial increase in PMF across all lower-limb muscles, with the largest relative changes observed in the tensor fasciae latae (118.86%), iliacus (115.57%), and gluteus maximus (101.37%). These results highlight the critical role of hip musculature in restoring dynamic stability following sudden deceleration events and complement previous electromyographic and biomechanical evidence regarding multi-muscle coordination during reactive balance control [8,9,13,22].

The observed increases in PMF reflect the functional demands of reactive recovery. Muscles in the high-response cluster, primarily responsible for proximal hip stabilization, exhibited force increases exceeding 100%, indicating that these muscles are essential for rapid restoration of balance and trunk stabilization. Moderate increases (40–100%) were seen in muscles contributing to hip and knee stabilization, as well as lower-limb support, suggesting that these muscles act to maintain limb alignment and facilitate weight transfer during the compensatory step. Distal and postural muscles displayed smaller increases (<40%), indicating a supportive role in fine-tuning postural control rather than serving as the primary drivers of the corrective response. These results support a hierarchical neuromuscular strategy in which proximal muscles generate the largest forces, followed by intermediate stabilizers, with distal muscles providing secondary postural support [8,9,22].

The prominent involvement of hip muscles aligns with prior studies. Liu and Lockhart [33] emphasized the critical role of hip joint responses to perturbations in the sagittal plane. Similarly, in combat sports, the hip muscles and proximal segments play a key role in generating striking power and postural control, as confirmed by kinetic and EMG analyses performed during mae-geri kicks in Kyokushin karate [34]. Similarly, Namayeshi, et al. [35] reported that the gluteus maximus plays a major role in recovery from perturbations applied during the initial contact phase. Shokouhi, et al. [36] also highlighted the importance of hip muscles, noting significant increases in positive work (concentric contractions) of hip musculature, although they identified the knee joint as the main contributor to negative leg work (eccentric energy absorption). In the present study, most knee muscles, except for the biceps femoris, were grouped in the low-response cluster, likely reflecting the different timing of perturbation application, which occurred during the PSw phase rather than immediately after heel strike.

The second cluster, showing moderate increases, mainly included muscles stabilizing the hip and ankle joints [37,38]. This underscores the importance of coordinated hip and ankle stabilization in response to sudden deceleration. Asghari, et al. [39] demonstrated that both hip and ankle stabilizing muscles play a vital role in balance recovery after gait perturbations, while Namayeshi, et al. [40] highlighted the critical contribution of ankle plantarflexors and the importance of their activation timing for successful recovery in young adults. The observed increases in PMFs of the soleus and tibialis posterior in the present paper further support the role of ankle muscles in reactive balance control.

The results also indicate that successful execution of the recovery step requires a muscle strength surplus across nearly all lower-limb muscles. This aligns with the conclusions of Sherrington, et al. [23], who emphasized the importance of resistance exercises in fall prevention programs for older adults, although the evidence was of moderate certainty. Notably, improvements in proprioception induced by resistance training [41] complicate the attribution of fall prevention solely to increased muscle strength. Moreover, some studies suggest that the ability to generate rapid muscle power may be more critical than maximal strength for effective perturbation recovery [42], highlighting the need for further research on the relative contributions of strength and power.

The use of static optimization in OpenSim allowed efficient estimation of individual muscle forces but introduced limitations, particularly regarding antagonist co-contraction and tendon compliance. Methodological evaluations suggest that static optimization may underestimate antagonist activity, potentially underrepresenting the forces generated by co-stabilizing muscles [43]. However, the relative PMF changes observed here are generally robust to these limitations and provide reliable insight into the direction and magnitude of muscle response adaptations.

Finally, the study population consisted of young, physically active females, providing insight into optimal neuromuscular control. Caution is warranted in generalizing these findings to older adults or clinical populations. Similarly, treadmill-based decelerations, while highly controlled, differ from overground perturbations in sensory feedback and task demands [44]. Nonetheless, split-belt perturbations are validated proxies for studying reactive stepping mechanics in repeatable laboratory conditions [44].

CONCLUSION

In summary, the recovery step following a treadmill belt deceleration in the pre-swing phase involves a graded, muscle-specific response, with proximal hip stabilizers generating the largest forces, intermediate stabilizers contributing to hip and ankle joint control, and distal or postural muscles providing fine-tuned postural support. These findings enhance our understanding of neuromuscular coordination during reactive balance and have important implications for designing targeted interventions to improve stability and reduce fall risk.