Wearable Visual Biofeedback of Vertical Ground Reaction Force Enables More Symmetrical Force Production During Deadlifting and Squatting

Abstract

Background/Objectives: Asymmetries in force production, characterized by vertical ground reaction forces (VGRFs), during lower-limb bilateral movements such as deadlifting and squatting, are considered biomechanical risk factors for injury. Real-time biofeedback has been used to modify lower limb force production but typically implements monitors. The purpose of this study was to determine the effect of wearable visual biofeedback (WVBF) on asymmetries in VGRFs and knee joint angles and the rate of perceived exertion (RPE) during deadlift and body-weight squatting (BWS) exercises in recreational powerlifters. Methods: Thirteen healthy young adults between 18–35 years of age performed three tasks: deadlifting for mixed-grip style (MIX), double-overhand style (DO), and BWS. Each task included two conditions: with and without WVBF. A two-way (Condition X Task) mixed model analysis of variance was performed to compare the bilateral asymmetry index of VGRFs, knee angle, and RPE scores. Results: A main effect of the condition (with versus without WVBF) was detected for VGRF symmetry (F (1,12) = 62.785, p < 0.001). WVBF showed decreased VGRF asymmetry compared to no biofeedback. For knee angle, a significant condition X task interaction (F (2,24) = 3.505, p < 0.05) was observed. For RPE, a main effect of the condition was observed (F (1,12) = 8.995, p < 0.05). WVBF showed greater RPE compared to no biofeedback. Conclusions: These results indicated that WVBF could reduce VGRF asymmetry during deadlifting and squatting. In addition, targeting force production symmetry may not directly yield joint angle symmetry and may increase perceived exertion. These results could provide valuable insight into VGRF modulation during deadlifting and squatting exercises in athletic and potentially clinical settings when targeting VGRF symmetry.

1. Introduction

Control of lower extremity force production is essential for bilateral dynamic movements such as deadlifts or squats. These demanding exercises challenge the musculature of the trunk and posterior chain including the glutes, hamstrings, and erector spinae [1] as well as the knee extensors and flexors [2]. However, squat-related movements are often accepted as symmetrical between the left and right sides of the body [3,4]. Still, up to 31% of sub-elite to elite powerlifters reported injuries during the deadlift exercise [5]. Excessive load, intensity, and poor technique were common causes of injuries in the lumbopelvic, hip, and shoulder regions for powerlifters [6,7]. In particular, symmetries in force production, characterized by the vertical ground reaction forces (VGRFs), are correlated with muscular strength imbalances [8], which should be mitigated through training to decrease poor performance [9,10] as such patterns could reinforce these imbalances during deadlifting [11]. Therefore, monitoring VGRF asymmetry may reduce risks of injury, functional asymmetries, and musculoskeletal imbalances. Coaches commonly utilize squat-related tasks to visually assess lower body flexibility and symmetry [12,13]. However, visually monitoring joint kinematics may not directly capture force asymmetry. Developing an approach that reduces force asymmetry during these exercises may provide a useful tool to reduce the risk of injury.

Real-time biofeedback has been used to improve mechanical performance and reduce asymmetries during limb-loading tasks in healthy and clinical populations when comparing control and feedback groups [14]. Previous studies have reported that task-specific feedback, such as frontal and sagittal plane-augmented feedback of self-performance (reviewing a video of prior performance and verbal instruction) during tuck jumping [15] and visual feedback on a tablet, of the center of pressure changes during one leg and tandem stance balance assessments have been an effective method for reducing biomechanical risk factors for injury and improving performance [16]. Applying real-time feedback of the peak propulsive force during level gait successfully allowed healthy younger and older adults to modulate lower extremity force production [17]. While these studies have used computer monitors, typically placed within 1 m of the participant, to display biofeedback [14,15,17], another study utilized a projector screen in front of a treadmill to display force biofeedback and reported a positive effect during knee extension angle exercises for individuals post anterior cruciate ligament injury [18]. In addition, a systematic review found that visual biofeedback has been shown to improve loading symmetry in transtibial amputees during gait and biofeedback systems and should be enjoyable for the user to motivate its long-term use [19]. The previous literature also suggests that clinical interventions such as visual biofeedback are needed to improve movement patterns and that load symmetry should be implemented prior to an athlete returning to play [20].

Traditional research utilizing biofeedback has typically implemented computer monitors, projector screens, or tablets to display visual force biofeedback [10,11,12,13,14,15,21,22,23]. This setup may require participants to constantly adjust gaze orientation to the monitor during dynamic movements, such as deadlifting and squatting. Previous research suggests that a downward gaze, when compared to an upward gaze, decreases the center of pressure displacement, further increasing stability during a back squat [24]. Postural deviations of the head and neck tilting back have been reported to increase the center of pressure displacement, resulting in poorer postural control [25]. These may be important aspects to consider when administering biofeedback during dynamic movements. This limitation could be resolved by using wearable headsets that can display force biofeedback directly to the users’ field of view. Previous studies have outlined the biomechanical advantages of virtual reality headsets to promote balance and gait in clinical populations [26], locomotor skill acquisition in virtual reality and transfer to real-world locomotion [27], and improved upper extremity exercise therapy in rehabilitation [28]. Therefore, wearable visual biofeedback (WVBF) may offer a more portable and functional approach to delivering biofeedback during deadlifts and squatting.

Deadlifting is typically performed using the mixed-grip style (MIX) over the double-overhand grip style (DO) as documented in the previous literature where 75% of males and 50% of females choose MIX rather than DO during the deadlift [29]. Although previous findings found grip style did not influence knee kinematics [30], the one-hand over one-hand under grip style generally provides more stability as it prevents the barbell from rolling in the lifter’s hands. However, to our knowledge, no study has examined how VGRF biofeedback is modulated during these two grip styles. Accordingly, we implemented WVBF during MIX, DO, and body-weight squatting (BWS). The purpose of this study was to determine the effect of WVBF on VGRFs and knee flexion/extension joint angle asymmetries during a deadlift and BWS exercise in recreational powerlifters. It was hypothesized that participants would exhibit more symmetrical force production and knee flexion/extension joint angle during the WVBF conditions compared to no WVBF.

2. Materials and Methods

Thirteen healthy younger adults (6 M; 7 F) between 18 and 35 years of age (22.46 ± 3.61 years) who lifted weights more than three times per week and had no diagnosed musculoskeletal injury in the past six months participated in this study. Participants’ body height (1.70 m ± 0.10 m), weight (80.07 kg ± 26.61 kg), and self-reported 1 repetition max (1 RM) for conventional deadlifts (101.61 kg ± 30.8 kg) were obtained. Participants self-reported years of experience deadlifting (4.53 ± 3.56 years). The dominant lower limb was determined by the answer to the question “If you were to kick a soccer ball for max distance, which limb would you use?”. All participants reported their right lower limb as the dominant limb. All participants provided informed consent and completed a physical activity readiness questionnaire 2023 (PAR-Q 2023) for a one-time visitation. The study hypothesis was not introduced to the participant at any point to avoid potential bias. This study was approved by the University of Texas at Austin Institutional Review Board.

Kinetic data were sampled at 1200 Hz using two force plates (Bertec, Columbus, OH, USA) to capture VGRF beneath each foot. Force plate data were filtered using a low-pass 4th order Butterworth filter with a cutoff frequency of 20 Hz and normalized to body weight in Visual3D (C-Motion Inc., Germantown, MD, USA). Kinematic data were captured using a 12-camera Vicon motion capture analysis system (Vicon, Oxford, UK) sampled at 120 Hz. Reflective marker placement corresponded to the Vicon Full-Body Plug-In-Gait Model [31]. Marker trajectories were filtered using a 4th order Butterworth low-pass filter at 6 Hz in Visual3D (C-Motion Inc., Germantown, MD, USA). A Meta Quest 2 (Meta, Menlo Park, CA, USA) was used during WVBF tasks to display real-time biofeedback of VGRFs during deadlift and BWS tasks with the passthrough mode enabled to allow subjects to see the lab environment and barbell beneath them through the headset. A custom Python (Python Software Foundation, Wilmington, DE, USA) code, along with Vicon DataStream SDK, was used to display real-time differences in VGRF data from two force plates, one beneath each foot, using the following equation:

$${\mathrm{F}\mathrm{z}}_{\mathrm{a}\mathrm{s}\mathrm{y}\mathrm{m}-\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{l}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}}={\displaystyle \frac{{\mathrm{F}\mathrm{z}}_{\mathrm{R}}-{\mathrm{F}\mathrm{z}}_{\mathrm{L}}}{{\mathrm{F}\mathrm{z}}_{\mathrm{R}}+{\mathrm{F}\mathrm{z}}_{\mathrm{L}}}}\times 100$$

(1)

where Fz_R and Fz_L are the real-time right and left VGRFs, respectively. The resultant differences, Fz_{asym−realtime}, were plotted and streamed to the Meta Quest 2 and displayed as a blue line for WVBF conditions (Figure 1). The highlighted region indicates the targeted margin of error: the top boundary represents 10% more body weight toward the right and bottom boundary represents 10% more body weight toward the left (totaling an allowable 20% margin of error). Bearing more weight on the right limb results in the VGRF difference being plotted above the region of symmetry. Bearing more weight on the left results in the VGRF difference plotted below the region of symmetry. VGRF differences plotted within the symmetry region indicate symmetric force production. This margin of error was determined by examining a previous study that found roughly 10% body weight sway unilaterally during quiet standing [32]. These data were streamed to the Meta Quest 2 via Zoom (Zoom, San Jose, CA, USA) with a refresh rate of 30 frames per second.

Participants were provided dark tight-fitted clothing to change into for data collection. Subject anthropometric data including height, weight, bilateral widths of ankle, knee, shoulder offset, widths of elbow, wrist, and hand thickness were measured along with the length from left after right anterior superior iliac crest to left and right medial malleolus to determine bilateral leg length. Retroreflective markers were applied to each subject. Subjects were instructed to stand with each foot on two separate force platforms, feet shoulder-width apart, and were presented with a barbell that weighed 40–50% (46.16 kg ± 12.92 kg) of their self-reported 1 RM for conventional deadlifts. Foot location was marked before exercise to ensure consistent foot placement throughout the experiment.

A self-selected warm-up period was provided for subjects before lifting. Subjects then performed three tasks under a control (without WVBF) condition: DO, MIX, and BWS. For each of the deadlift tasks, 1 set of 5 repetitions was performed, which is consistent with previous methodology [33]. Following the completion of each deadlifting set, RPE was recorded using the Borg scale [34]. For BWS, 10 consecutive squats were performed. A 3–5 min rest period was allowed between tasks until the participant gave verbal consent to continue. The order of the tasks was randomized. For the WVBF condition, participants were instructed to perform the same 3 tasks while receiving VGRF biofeedback (Figure 2), which consisted of biofeedback deadlift with double-overhand grip style (DO WVBF), biofeedback deadlift with mixed grip style (MIX WVBF), and biofeedback body weight squatting (BWS WVBF). Participants wore a Meta Quest 2 (Menlo Park, CA, USA) headset that displayed the differences in VGRF between the lower limbs. A familiarization period was provided to each subject to understand how to modulate their force production by shifting their weight to the left and right while performing each WVBF task. This consisted of practicing BWS and deadlifting tasks with WVBF for 5–10 min. Following the familiarization period, subjects performed the same three tasks in the same order when they first completed the control tasks. Participants only wore the virtual reality headset during the WVBF conditions.

Kinetic and kinematic data were calculated using Visual3D. Bilateral Asymmetry Index (BAI) scores were calculated for lower limb differences in VGRFs and knee flexion/extension joint angle using the same formula in Equation (1), which is consistent with the BAI calculation [35] where right and left refer to the right and left VGRF and knee flexion/extension angle variables. BAI scores closer to zero indicate symmetry and values further from zero indicate asymmetry between the limbs.

For all tasks, the average BAI scores for VGRFs and the knee flexion/extension angle were calculated for the entire duration of each repetition. The onset and offset of each repetition were determined by trough detection [36] on the cosine waveform in the vertical direction (z-axis) of the left-hand marker trajectory using a custom Python code. The first trough of the cosine waveform represented the onset point of the deadlifting repetition while the following trough represented the offset point of the same deadlifting repetition. The corresponding frame of the onset and offset time points were analyzed using a custom MATLAB (The MathWorks Inc., Natick, MA, USA) code, which calculated the maximum, minimum, average, range, and standard deviation of onset–offset windows identified through detection. The global average for bilateral VGRFs and knee flexion/extension joint angle values were extracted for each determined window and were used for BAI scoring.

A two-way (Condition X Task) mixed model analysis of variance was performed to compare BAI scores of the VGRF, knee flexion/extension angle, and RPE with and without WVBF. If a main effect was found, a Bonferroni post hoc analysis was performed for pairwise comparisons with an adjusted p-value of p < 0.0083. All statistical analyses were performed in Python 3.12 with a two-tailed alpha level of 0.05.

3. Results

3.1. Vertical Ground Reaction Forces

A main effect of the condition was detected on VGRF BAI (F (1,12) = 62.785, p < 0.001). Post-hoc analysis revealed that WVBF showed decreased VGRF asymmetry compared to the control (p < 0.001, mean difference: Control—WVBF = 8.205). There was no main effect of the task. A condition by task interaction was revealed (F (2,24) = 11.261, p = 0.001). Pairwise comparisons revealed that BWS tasks had greater mean differences in VGRF asymmetry across conditions when compared to all other tasks (p < 0.001, mean difference: MIX—MIX WVBF = 6.876, DO—DO WVBF = 6.665, BWS—BWS WVBF = 11.073) (Figure 3).

3.2. Knee Angle

There was a significant condition by task interaction on knee flexion/extension angle BAI (F (2,24) = 3.505, p < 0.046. Pairwise comparisons revealed that MIX tasks had a positive mean difference (mean difference: MIX—MIX WVBF = 1.551) whereas negative mean differences were revealed in DO and BWS tasks (mean difference: DO—DO WVBF = −0.550, BWS—BWS WVBF = −2.004). No main effect of the condition or task was revealed (Figure 4).

3.3. Rate of Perceived Exertion

For RPE, a main effect of the condition was observed (F (1,12) = 8.995, p < 0.05). Pairwise comparisons revealed significant negative mean differences between conditions (p < 0.05, mean difference: Control—WVBF = −1.115). A significant condition by task interaction was revealed (F (1,12) = 5.510, p < 0.05). MIX showed greater increases in RPE across conditions compared to DO (p < 0.05, mean differences: RPE MIX WVBF—RPE MIX = 1.692; RPE DO WVBF—RPE DO = 0.538). There was no main effect of the task.

4. Discussion

This study explored how WVBF affects VGRFs and knee flexion/extension joint angle symmetry in recreationally active healthy young adults during deadlifting and BWS tasks. Our results indicate that tasks with WVBF allowed subjects to produce more symmetrical VGRF production in all WVBF tasks. In addition, participants’ RPE increased during WVBF deadlift tasks. These findings indicate that providing WVBF of VGRFs may be an effective approach to promote VGRF symmetry during deadlifting and squatting exercises.

Compared to tasks without WVBF, all tasks involving WVBF promoted more symmetrical VGRF production during deadlifting and BWS (Figure 3). Our findings are consistent with prior research indicating that participants were able to effectively modulate their VGRF production during walking when provided with biofeedback [19]. Together, these findings support our hypothesis and elucidate the biomechanical effects of WVBF usage during deadlifting and BWS. Furthermore, BWS tasks revealed larger improvements with biofeedback in BAI scores when compared to MIX and DO tasks. It is possible that participants are more easily able to modulate VGRF distribution in response to the biofeedback due to no external load acting upon the body. Conversely, deadlift tasks involve increased demands in muscular strength, and in turn, could increase the difficulty of modulating functional force generation asymmetries [13]. Thus, future research that determines how different loads alter the effect of WVBF on force production could provide useful information to optimize the application of WVBF. Another potential explanation is that during the condition with no WVBF, the asymmetry score appeared to be greater during BWS compared to the weightlifting tasks and therefore allowed more room for correction during WVBF. Our findings expand from previous research that showed that ground reaction force biofeedback could increase force symmetry during gait [37,38] and improve standing balance [39] by demonstrating that this approach could be used during weightlifting and squatting exercises. A potential area for future investigation includes anterior cruciate ligament injury prevention programs where double-leg squatting is commonly employed [40,41]. The WVBF protocol could be further explored in this clinical population to reduce VGRF asymmetries that may be experienced during rehabilitation.

Our results in a healthy population demonstrated no significant differences in knee flexion/extension symmetry between control and WVBF tasks (Figure 4). While a condition-by-task interaction was detected, a deeper analysis excluding four statistical outliers revealed no significant interactions. Thus, the initial results of a significant condition by task interaction were driven by outliers in this sample. Importantly, knee flexion/extension angle symmetry was not the targeted form of biofeedback in the present study. A previous investigation found that knee kinetic and kinematic symmetry can improve in clinical populations when provided visual, verbal, and tactile biofeedback [42] and were successful in reducing knee hyperextensions during gait [43]. Therefore, knee flexion/extension angle and VGRF biofeedback each need to be specifically targeted as our results suggest that VGRF symmetry may not yield improved knee angle symmetry. Another potential explanation is that the BAI scores for the knee flexion/extension angles may be too low during the no WVBF condition and therefore restrict the effect of WVBF (floor effect). It is worth noting that to our current knowledge, no study has reported normal knee angle symmetry data, which warrants further investigations that outline joint kinematic symmetries during deadlifting.

Our results revealed deadlifting tasks involving WVBF that were significantly more exerting according to participants’ self-reported RPE using the Borg scale. A potential explanation could be that participants more frequently engaged musculature in both lower limbs during WVBF tasks to assist in VGRF modulation and increased physical exertion. Our results could be further explained by the novelty of WVBF on cognitive load. While RPE measures physical exertion only [44,45,46], it is possible that participants reported both combined physical and cognitive load in WVBF tasks. A previous dual-task study examined oxygen consumption when performing mental arithmetic problems on a bicycle ergometer and their results suggest that given an increase in cognitive demand, increases in metabolic requirements follow [47]. This may align with our findings since RPE increased during WVBF tasks given the participant’s focus on monitoring and modulating force production. However, future research should more closely examine cognitive demand and metabolic expenditure when paired with WVBF.

This study has limitations. First, this study had a small sample size of 13 recreationally active healthy young adults. Secondly, participants in this study only lifted 40–50% of their one RM and wore their desired shoes during data collection. Additional studies could examine how higher loads at 65–85% RM and/or during squat further challenge VGRF modulation while utilizing WVBF. Lastly, while RPE measures physical exertion, it could be possible that participants combined physical and cognitive fatigue in their RPE.

5. Conclusions

The present study showed that VGRF asymmetry could be improved with WVBF. In addition, targeting force production symmetry may not directly yield knee flexion/extension joint angle symmetry. The RPE results revealed a negative effect, possibly due to additional cognitive load and effort for force correction. The results of this study provide insight into how recreationally active populations respond to WVBF and could be directly translated into future rehabilitation and strength conditioning protocols when assessing lower extremity force production.