Sit-To-Walk Strategy Classification Using Hip and Knee Joint Angles at Gait Initiation

doi:10.21203/rs.3.rs-2718413/v1

Sit-to-walk (STW) is categorised by three movement strategies - forward continuation, balance and sit-to-stand-and-walk (SiStW). Literature identified strategies through biomechanical parameters using gold standard laboratory equipment, which is expensive, bulky, and not easily integrated into treatment solutions. As strategy becomes apparent at gait-initiation (GI) and the hip/knee are primary contributors in STW, this study proposes the hip/knee joint angles at GI, as an alternate and standalone method of strategy classification - measurable using wearable sensors. To achieve this, K-means clustering was implemented using three clusters and two feature sets (hip/knee angles); with data from an open access online database (age:21–80 years; n = 10). The results identified forward continuation with the lowest hip/knee extension at GI, followed by balance and then SiStW. From this classification, strategy biomechanics were investigated. The biomechanical parameters (derived in this study) that varied between strategies (P < 0.05) were time, horizontal centre of mass (COM) momentum, braking impulse, centre of pressure (COP) range and velocities, COP-COM separation, hip/knee torque and movement fluency. The derived strategy biomechanics are consistent with literature and validate the classification results. Through strategy classification an individual’s strategy-specific biomechanics can be understood and would aid the design and evaluation of interventions for movement impaired individuals.

Physical sciences/Engineering/Biomedical engineering

Health sciences/Anatomy/Musculoskeletal system

Health sciences/Health care/Quality of life

Sit-to-walk (STW) is a critical weight-bearing activity of daily living (ADL), with adults performing this task approximately sixty times daily¹. STW takes place when an individual transitions from a seated position to walking, via standing. This common motion, although seemingly basic, plays a major role in ensuring stability during time critical ADLs. For example, when we rush from a seated position to answer (1) a telephone, (2) a doorbell, or (3) when responding to an emergency. However, literature reports that the ability to execute STW deteriorates with age^2–4. Losing the ability to perform STW safely not only increases fall risk but also results in physical, psychological, and emotional degradation^5,6. Therefore, investigating and understanding the biomechanics, characteristics and execution of STW, is a vital first step in ensuring mobility, independent living and a good quality of life for adults and movement impaired individuals^2,4.

At present, there is numerous literature investigating sit-to-stand (SiSt). However, SiSt is merely a subset of STW, because it is normal to assume that an individual will ambulate immediately upon standing. Hence, in STW the end goal is walking, which is more common in daily life, making it a better representation of ADLs⁴. STW is defined as a fluid merging of SiSt and gait, at the point of gait-initiation (GI), where GI is denoted as the heel-off of the swing foot. Subsequently, the STW cycle can be divided into four transitionary phases: (1) flexion-momentum, (2) extension, (3) unloading, and terminating with (4) stance; after which, gait proceeds (Fig. 2)^2,7.

Literature reports of multiple studies that observed variations in STW biomechanics and executions (in their subject populations), and how these variations were generalised into different STW movement strategies. These identified strategies can be generally divided into three groups (Fig. 1): (a) Forward continuation, (b) Balance and (c) Sit-to-stand-and-walk (SiStW). Considering this, Magnan et al.,⁸ researched on the anteroposterior (AP) ground reaction forces (GRFs), centre of mass (COM) momentum and displacement, with centre of pressure (COP) trajectories, to differentiate between forward continuation and balance strategies in healthy adults. Similarly, Rousanoglou et al.,⁹ investigated movement speed and duration (fast and preferred speeds), COM velocity and displacement, COP trajectories and the temporal patterns of the STW transition phases, to also distinguish between such strategies. On the other hand, Bestaven et al.,¹⁰ and Buckley et al.,² considered the total COP trajectory, COM momentum, COP-COM separation and step length/velocity in relation to ageing to propose an alternate STW strategy commonly seen in older adults (SiStW). Furthermore, Chandler et al.,¹¹ and Kerr et al.,¹² studied the variation in movement fluency during STW, while Jones et al.,¹³ sought to find consistent biomechanical parameters between different STW strategies. Each STW strategy shows a different execution and can be described by a set of biomechanics, based on the variation in the above biomechanical parameters, investigated throughout literature.

As illustrated in Fig. 1, in forward continuation, a large horizontal COM (hCOM) momentum is generated, to propel the body forwards and upwards, while GI occurs earlier than in the other strategies (closer to seat-off). The feet or base of support (BOS) can be further away from the body (COM), as the generated momentum will carry the individual forwards. In balance, a braking impulse (posterior GRF) occurs to reduce the hCOM momentum generated and maintain quasi-static and postural stability, while rising. GI is delayed and the BOS is closer to the COM, compared to forward continuation^8,9. While in SiStW, a significant braking impulse occurs, with the BOS closest to the COM. This allows the individual to reach an almost upright position, before a delayed GI (compared to forward continuation or balance)². With this, Dehail et al.,¹⁴ showed that the quadriceps and hamstrings are the primary muscles involved in STW as they allow for hip and knee extension when rising, while modulating the braking impulse. Therefore, the hip and knee are primary contributors in STW.

The above literature highlighted the different executions of STW with their strategy-wise biomechanics, based on the investigated biomechanical parameters. To distinguish between the STW strategies all biomechanical parameters should be considered. However, literature lacked agreement on a single method of strategy classification. Additionally, to derive these biomechanical parameters, gold standard laboratory equipment like motion capture (Mocap) systems or force plates are required. Such equipment is expensive, bulky, cannot be easily integrated into treatment solutions, and not readily accessible in developing regions. Therefore, an alternative method of STW strategy classification using a single, standalone parameter is significant. This would enable strategy identification to be performed outside the laboratory, on wearable devices and in developing areas.

Based on the STW strategy executions (Fig. 1) and the definition of STW⁷, it is observed that the strategy first becomes apparent only at GI. This is because, before GI an individual begins rising symmetrically (SiSt) and the strategy is not yet distinguishable; however, after GI, as the swing foot moves forward, the chosen strategy is visible. At GI, the complete hCOM momentum and braking impulse generated are observable, along with the COM relative to the BOS, which determine the chosen strategy. Additionally, the hip and knee are primary contributors in STW¹⁴, with varying degrees of hip and knee joint extension, per strategy, at GI. As such, this study hypothesises that the hip and knee angles can serve as an alternative, distinguishing factor in classifying the STW strategy. Lower limb joint angles for strategy classification are beneficial as they are a standalone biomechanical parameter, independent of upper body movement, and can be easily measured using wearable sensors, in contrast to the biomechanical parameters from literature. Through STW strategy identification, an individual’s chosen STW execution biomechanics and characteristics can be described. This would aid treatment plans and interventions for movement impaired individuals, thus promoting independent living, easier access to ADLs and a better quality of life¹⁵. In this study, clustering was proposed to classify the strategies into forward continuation, balance and SiStW groups, based on the degree of hip/knee extension at GI. Furthermore, the varying STW strategy execution biomechanics and characteristics were investigated, to understand the strategies and for validation with literature.

Experiment details

Data from an open access database by Liang et al.,¹⁶ performed in the Rehabilitation Research Institute of Singapore, was considered in this study. The data collection was approved by the Nanyang Technological University Institutional Review Board (IRB-2018-04-014), and all subjects provided informed consent before commencement, in accordance with the Declaration of Helsinki. Raw Mocap and force plate data were provided in C3D format, from the NTU Dataverse database¹⁶. The dataset consisted of ten healthy subjects of Asian ethnicity (weight: 60.6 ± 11.3 kg, height: 166.5 ± 10.9 cm) with a wide age group ranging from 21 to 80 years, inclusive of all three STW strategies. Subjects performed the timed-up-and-go (TUG) test, as detailed in Chen & Chou¹⁷, with three repetitions, from which STW was obtained. During the TUG test, subjects were asked to stand from a seated position, walk forward for 3m, turn around, walk back, and sit down. Mocap data was obtained through a Qualisys (Sweden) Mocap system, sampling at 200 Hz, while GRF and COP data were obtained using two Kistler (Switzerland) force plates, sampling at 2000 Hz.

Data processing

First, the raw data was filtered to minimise noise, motion artifacts and for smoothing - using a zero-lag, second order, Butterworth lowpass filter with a cut-off frequency of 5 Hz and 20 Hz for Mocap and force plate data, respectively^18,19. The filter cut-off frequencies were selected by performing a Fast Fourier Transform and observing the 99% occupied signal bandwidth. Additionally, the force plate cut-off frequency was selected to preserve GRF events during GI.

Following this, OpenSim 4.2^20,21 was used for biomechanical analysis (see Supplementary Information), with STW being modelled using the Gait2392 Musculoskeletal Model²². This model was chosen as only the lower limbs were studied, while still accounting for upper body weight. Scaling was performed to match the subject’s anthropometry to the model, followed by inverse kinematics and inverse dynamics, to compute hip and knee joint angles and torques, respectively²³. Additionally, COM trajectories and velocities were derived through BodyKinematics analysis, while COP and GRFs were obtained directly from the force plates. Subsequently, these biomechanical parameters were analysed using MATLAB (Mathworks Inc.).

Data analysis

Joint angle clustering

As STW is defined as a fluid merging of SiSt and gait at the point of GI⁷, the three STW strategies become apparent only at GI. After GI, gait begins as the swing foot moves forward at ‘toe-off (TO) Swing’ (Fig. 2), with different execution mechanics. Based on the STW strategy executed, the hip and knee joint angles (which are the primary contributors of STW), vary at GI (Fig. 1), and thus can be used to identify the strategy.

K-means clustering was chosen for STW strategy classification as three distinctly identifiable strategies are known to exist, from literature. It is a fast and established algorithm; that can naturally identify these selected groups, within a numerical dataset with similar characteristics²⁴. As STW executions are grouped into three strategies, K-mean clustering was performed using three clusters and two data/feature sets - the hip and knee joint angles at GI^2,8. With this, Synthetic minority oversampling technique (SMOTE) was applied, to equalise sample length for each strategy. The cluster range was visualised using a circle, centred at the cluster centroid and radius equal to the Euclidean distance of the furthest point.

Considering this, the strategies were identified based on each cluster centroid’s degree of hip/knee extension at GI. The forward continuation cluster had the lowest hip/knee extension (largest joint angle magnitude) followed by balance (moderate hip/knee extension) and finally SiStW - which had the greatest hip/knee extension (lowest joint angle magnitude), as subjects were almost upright at GI. With this, silhouette analysis (describing cluster cohesion and separation) was performed, and it produced values greater than zero, showing that no data points were wrongly assigned to a cluster²⁵. Fuzzy C-means clustering was also performed using the hip and knee joint angles at GI and produced identical results to that from K-means clustering, thus validating the K-means clustering results.

Biomechanical parameters

A series of kinetic, kinematic and movement fluency parameters were derived, based on the four STW transition phases (Fig. 2). These parameters, as illustrated in Fig. 3, were analysed to understand the biomechanics of each of the three classified STW strategies in this study. The derived strategy-wise biomechanics were consistent with previously reported findings^{2,4, 8–10,12} and can be used to validate the joint angle based strategy classification.

The kinematic and kinetic parameters derived in this study were movement duration, hCOM and vertical COM (vCOM) momentum, braking impulse, COP range and velocities, COP-COM separation and joint angles and torques. Movement duration was obtained from the start of flexion-momentum (beginning of phase 1) to the end of stance (completion of phase 4), which terminates with a TO of the stance foot. Movement initiation was found using the first change in vertical GRF⁷, which also corresponds with the start of an anterior increase in hCOM velocity as the trunk flexes forward⁸. Next, the hCOM velocity at seat-off, represented the hCOM momentum generated during flexion-momentum, and the peak vCOM velocity represented the maximum momentum during extension. Additionally, the posterior GRF during flexion-momentum represented the braking impulse, where the area and peak were considered, as a percentage of bodyweight^8,9. Moreover, the AP and mediolateral (ML) COP instantaneous velocities and range were obtained as deviations from the force plate centre, while the AP/ML COP-COM separation was the difference between COM and COP trajectories. Both these quantities were obtained from extension to unloading phases because, during flexion-momentum there is no significant COP displacement as the feet remain on a single force plate, and during stance, the foot leaves the force plate anteriorly, with gait. Furthermore, the hip/knee joint angles and torques derived from OpenSim were found between seat-off and TO of the stance foot, at points of interest (Table 2). Joint angles were normalised to 0° when the subject was upright and for joint torque, with respect to the bodyweight. The maximum joint torques (not normalised), occurring around seat-off, were also tabulated.

Movement fluency reflects movement mechanical efficiency and comprises three objective measures - hesitation, coordination, and smoothness^11,12. Hesitation describes an uncertain movement and is the maximum percentage drop in hCOM velocity, from the initial peak before seat-off. Coordination describes the synchroneity of joint movements, via the percentage temporal overlap (with respect to total time) between when the hip/knee move into extension (C1 in Fig. 3) or flexion (C2 in Fig. 3) during the flexion-momentum and unloading phases, respectively. Smoothness describes motion inconsistency and was measured as the total number of inflections in the jerk hCOM signal.

Statistical analysis

In this work, SPSS Statistics (IBM) was used to conduct the statistical analysis; with tests using an α of 0.05. This analysis was conducted to find if a statistically significant difference existed between the biomechanical parameters (see the above section), for each STW strategy (Fig. 1); thus, allowing the biomechanics of each STW strategy to be investigated. The first analysis conducted was to test for normality using the Shapiro-Wilk test, which revealed that the distribution of the dataset was non-parametric. Following this finding, a Kruskal-Wallis H test was selected to find statistical differences between the three STW strategies (independent variable groups) for each biomechanical parameter (dependent variables). Subsequently, for parameters with a statistically significant difference (P < 0.05), a Mann-Whitney U test was performed as a post-hoc test, to find how each biomechanical parameter differed between the three strategies, and their extent.

Strategy classification

Table 1. Hip and knee joint angle ranges for each STW strategy cluster.

Hip and Knee Joint Angle Ranges for STW Strategy Clusters
Joint Angle Range (deg)	Forward Continuation		Balance		SiStW
Joint Angle Range (deg)	Hip	Knee	Hip	Knee	Hip	Knee
Maximum	84.32	76.63	68.70	59.74	41.45	25.89
Minimum	58.04	59.76	36.75	28.47	7.47	5.58

Figure 4 shows the K-means clustering results, which classify the three STW strategies using the hip/knee joint angles at GI, while Table 1 gives their respective joint angle ranges. From this illustration, forward continuation (red) had the lowest hip/knee extension (largest joint angle magnitude), followed by balance (green) and then SiStW (blue), which had the highest level of hip/knee extension as the subjects were almost upright at GI. An overlap exists between forward continuation and balance, showing a level of similarity and allowing them to be grouped together as previously reported by Magnan et al.⁸. Alternatively, SiStW acts as a separate postural conservative strategy which is consistent with the findings of Buckley et al.².

Strategy biomechanics

Table 2

Results of statistical analysis for kinematic, kinetic and movement fluency parameters
Biomechanical Parameters			(P < 0.05) for Kruskal-Wallis H Test	Mann-Whitney U Test (Post-hoc Testing)			Central Tendency and Variability - Median (IQR)
Biomechanical Parameters			(P < 0.05) for Kruskal-Wallis H Test	Forward & Balance	Forward & SiStW	Balance & SiStW	Forward	Balance	SiStW
Total time (s)			P < 0.001	0.003	P < 0.001	0.003	1.41 (0.20)	1.68 (0.38)	2.00 (0.13)
Horizontal COM velocity at seat-off (m/s)			P < 0.001	0.032	P < 0.001	0.001	0.57 (0.09)	0.46 (0.19)	0.35 (0.15)
Peak vertical COM velocity (m/s)			0.893	-	-	-	-	-	-
Braking impulse		Area (Ns)	P < 0.001	0.005	P < 0.001	0.002	2.40 (2.32)	4.46 (8.82)	15.63 (4.37)
Braking impulse		Peak (N)	0.025	0.028	0.015	0.488	55.28 (30.15)	98.35 (43.83)	86.04 (15.30)
COP range (mm)		AP	0.748	-	-	-	-	-	-
COP range (mm)		ML	P < 0.001	0.007	P < 0.001	0.001	84.22 (8.22)	113.62 (41.39)	164.14 (48.04)
Instantaneous COP velocity (m/s²)	Seat-off	AP	0.343	-	-	-	-	-	-
	Seat-off	ML	0.368	-	-	-	-	-	-
	Gait-initiation	AP	P < 0.001	0.954	P < 0.001	0.001	399.94 (124.50)	372.48 (219.33)	-95.79 (136.69)
	Gait-initiation	ML	P < 0.001	0.009	P < 0.001	0.011	-257.34 (153.97)	-548.54 (282.38)	-727.59 (225.19)
	Toe-off swing foot	AP	0.001	0.862	P < 0.001	0.002	343.01 (208.30)	307.61 (223.16)	-227.60 (155.65)
	Toe-off swing foot	ML	0.003	0.007	0.003	0.273	-386.48 (230.92)	-621.44 (246.14)	-711.57 (263.79)
COP-COM separation (mm)	Seat-off	AP	0.001	0.021	0.001	0.043	-185.25 (46.48)	-154.54 (21.67)	-141.21 (18.11)
	Seat-off	ML	0.010	0.954	0.038	0.002	281.79 (27.38)	276.92 (19.06)	264.84 (5.69)
	Gait-initiation	AP	0.111	-	-	-	-	-	-
	Gait-initiation	ML	0.679	-	-	-	-	-	-
	Toe-off swing foot	AP	0.107	-	-	-	-	-	-
	Toe-off swing foot	ML	0.178	-	-	-	-	-	-
Maximum joint torque (Nm)		Hip	0.061	-	-	-	146.63 (61.77)	133.23 (50.34)	189.32 (47.32)
Maximum joint torque (Nm)		Knee	0.152	-	-	-	94.58 (29.31)	84.14 (77.66)	139.10 (59.41)
Normalized joint torque (Nm/kg)	Seat-off	Hip	0.492	-	-	-	-	-	-
	Seat-off	Knee	0.274	-	-	-	-	-	-
	Gait-initiation	Hip	P < 0.001	0.012	P < 0.001	P < 0.001	1.68 (0.39)	1.13 (0.66)	0.59 (0.24)
	Gait-initiation	Knee	P < 0.001	0.001	P < 0.001	0.538	0.78 (0.39)	0.29 (0.64)	0.25 (0.19)
	Toe-off swing foot	Hip	P < 0.001	0.011	P < 0.001	0.001	1.08 (0.43)	0.72 (0.26)	0.42 (0.05)
	Toe-off swing foot	Knee	P < 0.001	0.002	0.024	P < 0.001	1.06 (0.41)	0.70 (0.53)	1.31 (0.17)
	Toe-off stance foot	Hip	0.007	0.003	0.069	0.094	0.61 (0.20)	0.45 (0.16)	0.52 (0.12)
	Toe-off stance foot	Knee	0.558	-	-	-	-	-	-
Movement fluency	Hesitation (%)		P < 0.001	0.024	P < 0.001	P < 0.001	13.56 (14.89)	26.34 (34.10)	64.74 (12.01)
	Coordination (%)	C1	P < 0.001	0.386	P < 0.001	0.001	-8.16 (2.03)	-7.29 (4.04)	-3.30 (2.42)
	Coordination (%)	C2	P < 0.001	0.47	P < 0.001	0.002	-3.99 (1.47)	-3.49 (0.80)	-2.51 (0.63)
	Jerk		0.002	0.155	0.001	0.017	14.00 (2.50)	15.50 (3.50)	18.00 (2.00)

Table 2 presents the statistical analysis of the derived biomechanical parameters. STW strategy classification through K-mean clustering (Fig. 4), was used when analysing these parameters. From these results, parameters with a statistically significant difference (P < 0.05) between the STW strategies (bolded P-values in Table 2), were used to describe the variation in biomechanics, for each individual strategy. This is significant, as understanding STW strategy biomechanics would aid the design and evaluation of treatment plans for movement impaired individuals.

Forward continuation

From Table 2, forward continuation recorded values of shortest duration (median: 1.41 s; IQR: 0.20), largest hCOM momentum (median: 0.57 m/s; IQR: 0.09) and lowest braking impulse (median area: 2.40 Ns; IQR: 2.32) amongst the three strategies. This results in a sharp trunk flexion, to rapidly propel the body forwards and upwards, yet requiring steady balance control to perform, without falling. Additionally, the large negative AP COP-COM separation at seat-off (median: -185.25 mm; IQR: 46.48) showed the COM lags the COP. This means the feet/base of support (BOS) is placed further away from the body (Fig. 1), while the large forward momentum keeps balance throughout the motion.

Thompson et al.,²⁶ showed that COP position and velocities act as balance predictors. Therefore, the lower ML COP range (median: 84.22 mm; IQR: 8.22) seen in forward continuation could show that this strategy was chosen by individuals with good balance control, such as healthy adults, which are consistent with the findings of Magnan et al.,⁸ and Rousanoglou et al.⁹. Moreover, Table 2 shows that forward continuation required greater hip and knee extension (lift) torque to raise the individual. This is due to the lower degree of joint extension as the individual is in a more crouched position when employing this strategy (Fig. 5).

Balance

From Table 2, the balance strategy recorded values of longer duration (median: 1.68 s; IQR: 0.38), lower hCOM momentum (median: 0.46 m/s; IQR: 0.19) and a higher braking impulse (median area: 4.46 Ns; IQR: 8.82), compared to forward continuation. The braking impulse reduces forward momentum and allows the individual to focus on stability and postural control, at the cost of speed and efficiency. Additionally, the lower AP COP-COM separation at seat-off (median: -154.54 mm; IQR: 21.67), shows the BOS closer to the COM (Fig. 1), allowing the individual to rise while maintaining quasi-static stability.

In contrast to forward continuation, the balance strategy had greater ML COP range (median: 113.62 mm; IQR: 41.39) which can indicate a more cautious STW execution⁸. Furthermore, Table 2 recorded lower hip and knee lift torque compared to forward continuation, as the subjects had a greater level of hip and knee extension or were in a less crouched position (Fig. 5).

Sit-to-stand-and-walk

SiStW reported values of longest duration (median: 2.00 s; IQR: 0.13), highest braking impulse (median area: 15.63 Ns; IQR: 4.37) and lowest hCOM momentum (median: 0.35 m/s; IQR: 0.15), compared to the other strategies (Table 2). The large braking impulse (resulting in lower hCOM momentum) allows the individual to stand, maintaining quasi-static stability and reach an almost upright position at GI. This is supported by the BOS being closest to the COM as shown by the lowest AP COP-COM separation (median: -141.21 mm; IQR: 18.11). These parameters minimise the fall risk, while maintaining good postural stability. Moreover, SiStW showed the highest ML COP range (median 164.14 mm; IQR: 48.04) and velocities, which may indicate uncertain movements from individuals with less stability and balance control and is consistent with the findings of Buckley et al.². However, tighter ML COP-COM separation at seat-off (median: 264.84 mm; IQR: 5.69) was observed, which keeps the feet/BOS in line with the COM, thus preserving movement stability^2,10. Therefore, SiStW allows individuals to maintain stability and balance during ADLs.

Hip and knee torque are progressively lower in SiStW (Table 2), as hip/knee extension is greater, and the individual is in a more upright position. As illustrated in Fig. 5, as the joints extend, the required torque decreases (from forward continuation to balance to SiStW). However, during TO of the swing foot, the knee flexes, as gait begins. Due to the larger knee extension in SiStW, the knee must move through a greater range of motion and higher knee flexion torque is required at this point.

Finally, movement fluency significantly differed (P < 0.05) between SiStW with the other strategies (Table 2), as similarly observed by Jones et al.¹³. For SiStW, a large braking impulse results in a greater hCOM momentum drop - greater hesitation and lower smoothness^11,12. Moreover, negative values and smaller magnitudes show weaker coordination as the knee moves into extension/flexion just before the hip. This further emphasises that SiStW is employed when an individual is uncertain of their motion, with weak motor control, balance, and stability.

STW is a vital ADL, therefore it is important to investigate and understand its biomechanics. Literature consisted of several studies that reported the variations in STW execution mechanics, which were generalised into three STW movement strategies^{2, 8–10,12,13}. All biomechanical parameters should be considered in strategy identification, hence literature lacked agreement on a single method of strategy classification. Additionally, these parameters required gold standard equipment (Mocap systems or force plates), which is bulky, expensive, and not readily integrated into treatment solutions. Therefore, an alternative method of STW strategy classification using the hip and knee joint angles at GI was proposed. This was chosen as the hip and knee are the primary contributors in STW¹⁴, while the strategy first becomes observable at GI. Strategy classification is important as it can be directly used to identify an individual’s STW biomechanics, based on their chosen strategy. This understanding would assist the design and evaluation of interventions in occupational therapy, for fall risk or movement impaired individuals; which in turn would improve quality of life.

K-means clustering was performed for STW strategy identification using the hip/knee joint angles, and these findings were used to analyse the variation in biomechanics (derived in this study) between each STW strategy (Table 2). The observed variation in STW execution biomechanics (for example with hCOM momentum or braking impulse) are consistent with the findings in literature describing the three STW strategies^{2, 8–10,12,13}. This demonstrates correct application of the K-means cluster grouping and thus, validates the use of hip/knee joint angles at GI, as an alternate method of STW strategy classification.

From literature, all biomechanical parameters need to be considered to reliably distinguish the STW strategy, however, joint angles are beneficial as they serve as a standalone parameter. The advantage of using joint angles is that they can be easily measured with wearable sensors. This allows it to be integrated into wearable interventions, used outside the laboratory, or even in developing areas that cannot facilitate such expensive laboratory setups. Additionally, lower limb joint angle strategy classification is independent of upper body movements - such as the use of arm swings to generate forward momentum.

An individual can employ either strategy at different instances, where the central nervous system selects and executes the best strategy. STW strategies and their biomechanics need to be considered when designing assistive devices, as they must support an individual to correctly perform their chosen strategy. Moreover, strategy classification could aid in the design and evaluation process of assistive devices. For such a device to be representative of STW, it should provide correct/sufficient assistance regardless of the employed strategy. Therefore, through strategy classification the assistive device can be evaluated against each of the three strategies. Only if the device can perform for all three STW strategies, would it encompass the entire STW motion - describing its efficacy, effectiveness, and applicability to STW.

On that account, lift assistive devices should consider the strategy-wise torque variation, to ensure correct levels of assistive torque are provided. Forward continuation required the largest extension torque, followed by balance. In contrast, SiStW required less extension torque, but greater knee flexion torque at TO. This result, coupled with lower hCOM momentum, tighter AP and ML COP-COM separation, and a larger braking impulse, showed that SiStW had a lower fall risk during ADLs. Therefore, SiStW is preferred amongst movement impaired individuals, such as older adults², while balance and forward continuation require better postural and balance control as characteristic of healthy adults⁸.

A limitation of this study is only considering torque at points of interest and not investigating the overall STW torque profile, per strategy. Furthermore, the study did not investigate arm strategies⁴, asymmetric foot position or the use of walking aids as commonly used in ADLs. This could affect the momentum generated and balance control and will be researched in future studies.

This study proposed using the hip/knee joint angles at GI, as an alternate method of distinguishing between the three STW strategies - forward continuation, balance and SiStW. Based on this grouping, the strategy-wise biomechanics were derived, and are consistent with existing literature, thus validating this method of strategy identification. The biomechanical parameters governing STW strategies are hCOM momentum, braking impulse, ML COP range, COP-COM separation at seat-off, joint torque and movement fluency. These strategy biomechanics, coupled with strategy classification would aid the evaluation and design of treatment plans and interventions. Failing to do so, would increase fall risk and impede access to ADLs, for movement impaired individuals.

ADL:	Activity of daily living
AP:	Anteroposterior
COM:	Centre of mass
COP:	Centre of pressure
GI:	Gait initiation
GRF:	Ground reaction force
hCOM:	Horizontal centre of mass
IQR:	Interquartile range
ML:	Mediolateral
Mocap:	Motion capture
SiSt:	Sit-to-stand
SiStW:	Sit-to-stand-and-walk
SMOTE:	Synthetic minority oversampling technique
STW:	Sit-to-walk
TO:	Toe-off
TUG:	Timed-up-and-go
vCOM:	Vertical centre of mass

Competing interests

The authors declare no competing interests.

Data availability

All datasets analysed during this study are included in the published article by Liang et al.,¹⁶ and can be found under the following hyperlink: https://doi.org/10.1038/s41597-020-00627-7

Acknowledgements

This work is supported by the Ministry of Higher Education, Malaysia under the project number: FRGS/1/2022/TK07/MUSM/02/2

Author contributions

CP and AG conceived the study and the extent of its scope. CP conducted the study under the supervision of AG. The results for the study were obtained by CP, while the interpretation of the results and their respective conclusions were formulated by CP and AG with support from DG and SA. The manuscript was drafted by CP and AG and reviewed by DG, SA and MS. All authors read and approved the final manuscript.

Dall, P. M. & Kerr, A. Frequency of the sit to stand task: An observational study of free-living adults. Appl Ergon 41, 58–61 (2010).
Buckley, T., Pitsikoulis, C., Barthelemy, E. & Hass, C. J. Age impairs sit-to-walk motor performance. J Biomech 42, 2318–2322 (2009).
Robinovitch, S. N. et al. Video capture of the circumstances of falls in elderly people residing in long-term care: an observational study. Lancet 381, 47–54 (2013).
van der Kruk, E. et al. Why do older adults stand-up differently to young adults?: investigation of compensatory movement strategies in sit-to-walk. npj Aging 8, 1–19 (2022).
Pozaic, T., Lindemann, U., Grebe, A.-K. & Stork, W. Sit-to-Stand Transition Reveals Acute Fall Risk in Activities of Daily Living. IEEE J Transl Eng Health Med 4, 2700211–2700211 (2016).
Xie, H., Chen, P. W., Zhao, L., Sun, X. & Jia, X. J. Relationship between activities of daily living and depression among older adults and the quality of life of family caregivers. Frontiers of Nursing 5, undefined-undefined (2018).
Kerr, A., Durward, B. & Kerr, K. M. Defining phases for the sit-to-walk movement. Clin Biomech (Bristol, Avon) 19, 385–390 (2004).
Magnan, A., McFadyen, B. J. & St-Vincent, G. Modification of the sit-to-stand task with the addition of gait initiation. Gait & Posture 4, 232–241 (1996).
Rousanoglou, E. N., Kondilopoulos, N. & Boudolos, K. D. Fast Motion Speed Alters the Sit-to-Walk Spatial and Temporal Pattern in Healthy Young Men. Sports Med Int Open 4, E77–E84 (2020).
Bestaven, E., Petit, J., Robert, B. & Dehail, P. Center of pressure path during Sit-to-Walk tasks in young and elderly humans. Annals of Physical and Rehabilitation Medicine 56, 644–651 (2013).
Chandler, E. A. et al. Investigating the Relationships Between Three Important Functional Tasks Early After Stroke: Movement Characteristics of Sit-To-Stand, Sit-To-Walk, and Walking. Front Neurol 12, 660383 (2021).
Kerr, A., Pomeroy, V. P., Rowe, P. J., Dall, P. & Rafferty, D. Measuring movement fluency during the sit-to-walk task. Gait & Posture 37, 598–602 (2013).
Jones, G. D., James, D. C., Thacker, M. & Green, D. A. Parameters that remain consistent independent of pausing before gait-initiation during normal rise-to-walk behaviour delineated by sit-to-walk and sit-to-stand-and-walk. PLOS ONE 13, e0205346 (2018).
Dehail, P. et al. Kinematic and electromyographic analysis of rising from a chair during a “Sit-to-Walk” task in elderly subjects: Role of strength. Clinical Biomechanics 22, 1096–1103 (2007).
Norman-Gerum, V. & McPhee, J. Comprehensive description of sit-to-stand motions using force and angle data. Journal of Biomechanics 112, 110046 (2020).
Liang, P. et al. An Asian-centric human movement database capturing activities of daily living. Sci Data 7, (2020).
Chen, T. & Chou, L.-S. Effects of Muscle Strength and Balance Control on Sit-to-Walk and Turn Durations in the Timed Up and Go Test. Arch Phys Med Rehabil 98, 2471–2476 (2017).
Crenna, F., Rossi, G. B. & Berardengo, M. Filtering Biomechanical Signals in Movement Analysis. Sensors (Basel) 21, 4580 (2021).
Winter, D. Biomechanics and Motor Control of Human Movement, Fourth Edition. (2009) doi:10.1002/9780470549148.ch5.
Delp, S. L. et al. OpenSim: open-source software to create and analyze dynamic simulations of movement. IEEE Trans Biomed Eng 54, 1940–1950 (2007).
Seth, A. et al. OpenSim: Simulating musculoskeletal dynamics and neuromuscular control to study human and animal movement. PLOS Computational Biology 14, e1006223 (2018).
OpenSim. Musculoskeletal Models - OpenSim Documentation. https://simtk-confluence.stanford.edu/display/OpenSim/Musculoskeletal+Models (2021).
OpenSim. User’s Guide - OpenSim Documentation. https://simtk-confluence.stanford.edu/display/OpenSim/User%27s+Guide (2021).
MATLAB-Kmeans. k-means clustering. https://www.mathworks.com/help/stats/kmeans.html (2021).
Shutaywi, M. & Kachouie, N. N. Silhouette Analysis for Performance Evaluation in Machine Learning with Applications to Clustering. Entropy (Basel) 23, 759 (2021).
Thompson, L. A., Badache, M., Cale, S., Behera, L. & Zhang, N. Balance Performance as Observed by Center-of-Pressure Parameter Characteristics in Male Soccer Athletes and Non-Athletes. Sports (Basel) 5, 86 (2017).

No competing interests reported.

Supplementaryinformation.pdf

Sit-To-Walk Strategy Classification Using Hip and Knee Joint Angles at Gait Initiation

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Methods

Experiment details

Data processing

Data analysis

Joint angle clustering

Biomechanical parameters

Statistical analysis

Results

Strategy classification

Strategy biomechanics

Forward continuation

Balance

Sit-to-stand-and-walk

Discussion

Conclusion

Abbreviations

Declarations