Sensor-Based Analysis of Upper Limb Motor Coordination After Stroke: Insights from EMG, ROM, and Motion Data During the Wolf Motor Function Test

Abstract

The Wolf Motor Function Test (WMFT) is widely used to evaluate upper limb motor performance after stroke. However, conventional approaches may overlook domain-specific neuromuscular and kinematic differences during task execution. This study classified WMFT tasks into three functional domains: proximal reaching and transport (PRT), fine motor manipulation (FMM), and gross motor functional control (GMFC). Interlimb differences in muscle activation, joint mobility, and movement amplitude were examined using sensor-based measurements. Twelve individuals with chronic stroke performed 16 WMFT tasks. Surface electromyography (EMG) and inertial measurement units (IMUs) recorded upper limb muscle activity, joint angles, and segmental displacement. Wilcoxon signed-rank tests and Spearman correlations were conducted for each functional domain. Significant asymmetries in EMG, range of motion (ROM), and root mean square (RMS) acceleration were found in PRT and FMM tasks. These results reflect increased proximal muscle activation and reduced distal engagement on the paretic side. GMFC tasks elicited more symmetrical patterns but still showed subtle deficits in distal control. Correlation analyses demonstrated strong interdependencies among neuromuscular and kinematic measures. This finding underscores the integrated nature of compensatory strategies. Categorizing WMFT tasks by functional domain and integrating multimodal sensor analysis revealed nuanced impairment patterns. These patterns were not detectable by conventional observational scoring. These findings support the use of sensor-based, domain-specific assessment to guide individualized rehabilitation strategies. Such approaches may ultimately enhance long-term functional recovery in stroke survivors.

1. Introduction

Stroke is among the most prevalent neurological conditions affecting older adults and remains a leading cause of long-term disability worldwide. More than 80% of stroke survivors develop hemiparesis, which often results in impaired upper limb function. This impairment leads to difficulties in performing essential activities of daily living (ADLs), such as reaching, grasping, and manipulating objects [1].

Upper limb motor impairments following stroke commonly manifest as muscle weakness, spasticity, joint stiffness, and loss of selective motor control, particularly in the contralesional side of the body. These deficits hinder coordinated movement and reduce independence in daily tasks such as feeding, dressing, and writing. To compensate, many stroke patients adopt maladaptive movement patterns, including excessive trunk rotation, shoulder elevation, and abnormal muscle synergies. Although these compensations may temporarily assist in task completion, they often reinforce inefficient motor control and impose excessive strain on other joints and muscles. Early detection of such patterns, whether by clinical observation or quantitative motion analysis, enables targeted interventions. Timely intervention can prevent the reinforcement of maladaptive behaviors. It also facilitates motor relearning and increases the likelihood of regaining long-term functional independence [2,3]. However, even in the long-term chronic phase, often more than a decade post-stroke, patients may continue to exhibit persistent motor deficits. These individuals frequently demonstrate entrenched compensatory strategies. Investigating this population is critical for understanding stable movement patterns that evolve over years. Such patterns may necessitate rehabilitation approaches distinct from those applied in the acute or subacute stages.

Because the upper limb plays an essential role in interacting with the environment, restoring its function remains a key objective in neurorehabilitation. However, conventional clinical assessments often rely on time-based or observational scales. These methods lack the sensitivity to detect subtle motor deficits or compensatory strategies employed during functional tasks [4]. Consequently, recent studies have highlighted the importance of sensor-based assessments. Methods such as inertial measurement units (IMUs) and surface electromyography (EMG) enable the capture of real-time neuromuscular and kinematic data during upper limb activity [5].

The Wolf Motor Function Test (WMFT) is a widely used clinical tool that evaluates upper limb performance through standardized functional tasks. Although clinically useful, the WMFT primarily yields time scores and ordinal ratings. These measures provide limited insight into underlying motor control strategies or quality of movement [6,7]. Several studies have shown that stroke patients often complete WMFT tasks using atypical strategies such as joint substitution, trunk compensation, or synergistic movements. Such deviations may be overlooked with conventional scoring systems [2,8,9]. Furthermore, these scores fail to capture movement smoothness, coordination, or joint-specific contributions to task performance [10]. As a result, this limitation can lead to overestimation of recovery and suboptimal treatment planning.

To address these challenges, there is increasing interest in augmenting the WMFT with wearable sensor technologies. By incorporating EMG and IMU data, researchers can analyze specific movement features in real time. These include joint range of motion (ROM), acceleration profiles, and muscle activation patterns. This multimodal approach provides deeper sight into motor control and compensation patterns at the individual level [11,12,13].

Upper limb movements vary depending on whether a task requires proximal transport, fine motor manipulation, or gross motor control. Each task type engages distinct joints, muscle groups, and control mechanisms. However, few studies have systematically categorized WMFT tasks by functional domain or examined sensor-based data across these categories. Recent research has begun to differentiate gesture-based from grasp-based movements. Significant differences have been observed in trunk involvement, shoulder–elbow coordination, and motion smoothness [14]. Wearable sensors have also enabled movement segmentation based on IMU signals. In addition, advances in coordination analysis have demonstrated that task-specific movement strategies can be identified [15,16]. Despite these developments, systematic reviews of robot-assisted upper limb rehabilitation note that task-level distinctions are seldom reported. This gap continues to limit progress in personalized rehabilitation [17].

This study conducted a preliminary quantitative analysis of upper limb function in stroke patients during WMFT tasks. The tasks were categorized into three functional domains: proximal reaching and transport (PRT), fine motor manipulation (FMM), and gross motor functional control (GMFC). This categorization is consistent with prior kinematic and neurorehabilitation frameworks [18]. These frameworks distinguish transport-oriented proximal movements from manipulation-oriented distal movements. This framework is grounded in motor control theory. The theory differentiates between proximal transport components and distal manipulation components of goal-directed actions. It also identifies gross stabilizing movements that support these actions. Proximal tasks such as reaching and transporting primarily engage shoulder and elbow control systems. After stroke, these systems are often affected by compensatory strategies involving the trunk and shoulder. In contrast, distal manipulation tasks require refined hand and finger control. This control is frequently compromised due to impaired selective motor activation [19,20]. GMFC tasks primarily demand gross coordination and postural stability. This task category places comparatively less emphasis on fine dexterous control compared to PRT and FMM tasks [21]. Traditional WMFT scoring aggregates all task performances into a single score. This approach may limit the detection of selective deficits in proximal stability, distal dexterity, or whole-limb coordination. By analyzing these functional domains separately, our approach enables identification of domain-specific neuromuscular and kinematic impairment patterns. Such patterns are often not detected in aggregate scoring. These distinctions have been used in previous research to guide targeted rehabilitation interventions. This categorization also enables more sensitive monitoring of recovery trajectories. Building on this framework, we integrated EMG, ROM, and IMU-derived kinematic measures to compare performance between the paretic and non-paretic limbs. The comparison focused on muscle activation, joint mobility, and movement amplitude.

Based on the prior literature and clinical observations, we hypothesized that

(H1). 

PRT and FMM tasks would elicit greater asymmetries than GMFC tasks. These asymmetries would be characterized by excessive recruitment of proximal stabilizers and reduced activation of distal muscles.

(H2). 

GMFC tasks would exhibit relatively symmetrical patterns, while still indicating subtle deficits in distal motor control.

(H3). 

Movement amplitude deficits would correlate strongly with both proximal and distal impairments. This relationship is indicated by associations among EMG, ROM, and RMS measures.

These hypotheses aimed to identify domain-specific neuromuscular and kinematic features of stroke-related hemiparesis. The findings are intended to inform individualized rehabilitation strategies. This evidence also supports the development of real-time, sensor-based monitoring systems for more objective clinical assessment.

2. Materials and Methods

2.1. Participants

Twelve right-handed individuals (nine males and three females) were recruited for this study. All participants were aged 65 years or older and had been diagnosed with right hemiplegia following a stroke. Only right-handed participants were included to reduce variability related to hand dominance. Dominant hand use can influence both pre-stroke motor patterns and post-stroke compensatory strategies. Therefore, restricting the sample to right-handed individuals ensured more consistent interpretation of motor impairment. This restriction allowed clearer comparisons across participants. All participants were receiving occupational therapy at I Community Rehabilitation Center following diagnosis at W Hospital in Jeonbuk State. Inclusion criteria were as follows: (1) stroke onset more than six months prior to participation; (2) sufficient cognitive ability to understand and follow instructions; and (3) no history of other neurological or orthopedic conditions unrelated to stroke. The mean age, height, and body weight of participants were 71.3 ± 6.6 years, 166.2 ± 9.2 cm, and 72.4 ± 7.9 kg, respectively. The mean disease duration since stroke onset was 13.70 ± 7.66 years. This indicates that all participants were in the long-term chronic phase of recovery. The cohort was evenly divided in terms of stroke type. Six participants had a hemorrhagic stroke, and six had an ischemic stroke. Before participation, all individuals received detailed information about the study’s purpose and procedures. Written informed consent was then obtained. The study protocol was reviewed and approved by the Institutional Review Board of Jeonbuk National University.

2.2. Experimental Procedure

Patients’ upper limb function was assessed using the WMFT. This standardized and validated tool is commonly employed to assess motor performance in individuals recovering from stroke [22,23]. The WMFT is designed to measure both the time and functional ability required to complete a series of upper extremity tasks. It provides clinically relevant information on the level of motor recovery and residual impairment [24]. The test has been validated for its reliability, internal consistency, and sensitivity to changes in motor performance after stroke [25,26]. The WMFT includes 17 items: 15 time-based tasks that reflect daily upper limb functions, such as reaching, lifting, and fine motor manipulation, and 2 strength-based tasks (weight to box and grip strength). These tasks are performed sequentially. Scoring is based on both completion time and a functional ability scale ranging from 0 to 5, with higher scores indicating better performance.

In this study, all WMFT tasks were performed in a seated position with standardized task setups to maintain consistency across participants. As recommended in clinical protocols, assessments began with the non-paretic limb to reduce fatigue and control for learning effects. This procedure facilitated clearer comparison of motor performance between the paretic and non-paretic sides. It also enabled the identification of functional asymmetries and compensatory strategies. The grip strength task, which primarily evaluates static force rather than dynamic movement, was not included in the sensor-based analysis. In contrast, although the weight-to-box task is traditionally categorized as a strength-based assessment, it was treated in this study as a movement-based task. This reclassification reflects its dynamic components, including reaching, load transfer, and object placement [27]. The remaining 16 items were classified into three functional domains based on joint involvement and motor control demands: proximal reaching and transport (PRT), fine motor manipulation (FMM), and gross motor functional control (GMFC). This categorization enabled task-specific comparisons of muscle activation, movement amplitude, and joint mobility between limbs, thereby allowing for more targeted analysis of functional performance [28].

Table 1. Classification of WMFT tasks by functional domain.

Functional Domain	Task Name
Proximal Reaching and Transport (PRT)	Forearm to Table (side)
	Forearm to Box (side)
	Extend Elbow (side)
	Extend Elbow (with weight)
	Hand to Table (front)
	Hand to Box (front)
	Reach and Retrieve
	Lift Basket
	Weight to Box
Fine Motor Manipulation (FMM)	Lift Pencil
	Lift Paper Clip
	Stack Checkers
	Flip Cards
	Turn Key in Lock
Gross Motor Functional Control (GMFC)	Lift Can
	Fold Towel

summarizes the categorization of the 16 analyzed WMFT tasks into the three functional domains: PRT, FMM, and GMFC. The classification is based on joint involvement and motor control demands, enabling domain-specific comparison of neuromuscular and kinematic performance.

Figure 1 illustrates the bilateral placement of EMG electrodes and IMU sensors on anatomically relevant landmarks. This setup allows synchronized recording of muscle activity and segmental motion. Upper limb muscle activity was measured using a wireless surface EMG system (Noraxon Desktop DTS, Noraxon Inc., Scottsdale, AZ, USA) during the performance of WMFT tasks [29,30]. EMG signals were sampled at 1500 Hz. Surface electrodes were bilaterally attached to eight muscles. These included the upper trapezius (UT), anterior deltoid (AD), triceps brachii (TB), biceps brachii (BB), extensor carpi radialis longus (ECRL), pronator quadratus (PQ), flexor carpi ulnaris (FCU), and flexor carpi radialis (FCR). These muscles were selected to assess their major functional roles. Specifically, they contribute to shoulder stabilization and elevation (UT), shoulder flexion (AD), elbow flexion and extension (BB, TB), wrist flexion and extension (FCR, FCU, ECRL), and forearm pronation (PQ). All of these functions are essential for executing the reaching, lifting, and manipulation tasks included in the WMFT [31,32]. Prior to electrode placement, the skin over each muscle belly was cleaned with alcohol to minimize impedance and improve signal quality. Electrodes were positioned according to SENIAM recommendations to ensure optimal signal acquisition and reduce crosstalk. Muscle activity was recorded separately for the paretic side (PS) and non-paretic side (NPS), thereby allowing direct comparison of neuromuscular control during upper limb task execution in stroke patients [10].

Joint-specific ROM and overall movement amplitude during WMFT tasks were assessed using a wearable inertial motion capture system (Noraxon MyoMotion Research, Noraxon Inc., Scottsdale, AZ, USA). This system incorporates IMUs that combine accelerometers, gyroscopes, and magnetometers to provide three-dimensional orientation and displacement data in real time. Seven IMU sensors were positioned on the following anatomical landmarks: head, upper thoracic spine, lower thoracic spine, pelvis, upper arms, forearms, and hands. These placements enabled assessment of multi-joint coordination and segmental kinematics across the upper body during functional upper limb tasks. Before sensor attachment, the skin over each anatomical site was cleaned with alcohol wipes to reduce movement artifacts and maintain stable sensor adhesion. Calibration was performed in a standardized upright posture according to the manufacturer’s protocol, establishing a neutral reference frame for kinematic analysis. For each participant and each WMFT task, data were collected for three valid trials per limb. The resulting EMG and IMU parameters were averaged to yield a representative value for analysis.

2.3. Data Analysis

EMG and IMU sensor data collected during WMFT performance were analyzed for comparison of upper limb performance between the PS and NPS across the three functional domains. Upper limb muscle activity was assessed using the MyoResearch3 (MR3) version 3.6 software (Noraxon Inc., Scottsdale, AZ, USA). Raw EMG signals were first band-pass filtered (20–450 Hz) to attenuate movement artifacts and powerline noise [31]. A 60 Hz notch filter was then applied to suppress electrical interference. The filtered signals were subsequently full-wave rectified and smoothed using a 100 ms moving root mean square (RMS). Muscle activation data were normalized using the percentage reference voluntary contraction (%RVC) method. The %RVC serves as an indicator of upper limb muscle recruitment efficiency and was derived from the peak activation value of each muscle recorded during task performance [33].

IMU-derived joint kinematic data were processed using Noraxon’s software suite to extract joint angles and segmental displacements during each WMFT task. ROM was defined as the difference between the maximum and minimum joint angles measured during task performance [34]. The analysis focused on ROM variables relevant to upper limb function and EMG electrode placement. These included shoulder flexion (SF), shoulder abduction (SA), shoulder external rotation (SER), elbow flexion (EF), wrist flexion (WF), wrist radial deviation (WRD), wrist supination (WS), upper arm pitch (UAP), upper arm roll (UAR), and forearm external rotation (FER). Variables were selected for relevance to upper limb kinematics and correspondence with major joints and movement planes involved in reaching, lifting, and manipulation. Specifically, SF, SA, and SER assessed multi-planar mobility of the glenohumeral joint. EF reflected positioning and transport capacity. WF, WRD, and WS represented fine motor and grasping functions. UAP, UAR, and FER captured segmental coordination and rotational control during task performance. To improve clarity in reporting and interpretation, ROM variables were grouped into functional categories: shoulder movements (SF, SA, SER), elbow motion (EF), wrist movements (WF, WRD, WS), upper arm orientations (UAP, UAR), and forearm rotation (FER). In addition to ROM, segmental movement characteristics were quantified using IMU data. The root mean square (RMS) of the upper arm, forearm, and hand was calculated across the X (medio-lateral), Y (antero-posterior), and Z (vertical) axes. RMS values were obtained from accelerometer signals using sensor fusion algorithms [35]. This analysis focused on three anatomical segments and axes to capture multi-directional displacement characteristics during task performance. This approach yielded a comprehensive measure of spatial movement amplitude, corresponding to lateral (X), forward–backward (Y), and vertical (Z) motion during task performance. Although task completion time is a standard WMFT metric, the analysis prioritized sensor-based measures of muscle activation, joint mobility, and movement amplitude. For each task, the average ROM and RMS values were computed separately for the PS and NPS. This allowed for detailed comparisons of joint flexibility, movement magnitude, and segmental motor control. For ease of interpretation, abbreviations were used to denote each segment and axis combination: UAX (Upper Arm X), UAY (Upper Arm Y), UAZ (Upper Arm Z), FAX (Forearm X), FAY (Forearm Y), FAZ (Forearm Z), HX (Hand X), HY (Hand Y), and HZ (Hand Z).

Given the relatively small sample size and the number of statistical tests performed, the Bonferroni method was used to adjust for multiple comparisons. This approach controlled the familywise Type I error rate in pairwise comparisons between the PS and NPS. Both adjusted and unadjusted p-values are reported. Furthermore, effect sizes were calculated for all Wilcoxon signed-rank tests to quantify difference magnitudes independent of sample size. The Shapiro–Wilk test was used to assess data normality. As normality was not consistently satisfied, the Wilcoxon signed-rank test was applied to compare EMG, ROM, and RMS values between the PS and NPS. Spearman’s rank correlation was performed to evaluate the relationships among muscle activation, joint mobility, and movement amplitude. The analysis tested the hypothesis that impaired neuromuscular activation would be associated with reduced joint mobility and altered segmental displacement patterns. For each variable, 95% confidence intervals for the mean values were estimated using a bootstrap resampling approach with 5000 iterations. This procedure quantified the precision of the point estimates. Statistical significance was set at p < 0.05 after Bonferroni adjustment, with p < 0.01 and p < 0.001 denoting increasing strength of evidence against the null hypothesis.

3. Results

3.1. Muscle Activation

Table 2. Differences in muscle activation (% of RVC) of the upper limbs between the paretic side (PS) and non-paretic side (NPS) during proximal reaching transport (PRT), fine motor manipulation (FMM), and gross motor functional control (GMFC) tasks of the Wolf Motor Function Test (WMFT), with Bonferroni-adjusted p-values and effect sizes (r).

Muscle		PS (95% CI)	NPS (95% CI)	p-Value	Adjusted p (Bonferroni)	r
PRT	UT	38.06 ± 9.03 (35.45–40.19)	32.25 ± 8.50 (29.79–34.35)	<0.001	<0.001 ***	0.553
	AD	31.57 ± 8.96 (29.18–33.73)	30.20 ± 7.61 (28.04–31.94)	0.556	1.000	0.070
	TB	19.33 ± 4.88 (17.97–20.52)	20.95 ± 5.25 (19.52–22.14)	0.217	1.000	0.148
	BB	26.50 ± 8.13 (24.49–28.51)	26.49 ± 6.65 (24.63–28.07)	0.602	1.000	0.062
	ECRL	28.07 ± 7.73 (25.97–29.99)	37.55 ± 5.36 (35.79–38.73)	<0.001	<0.001 ***	0.878
	PQ	19.20 ± 5.80 (17.62–20.60)	27.38 ± 6.44 (25.66–28.95)	<0.001	<0.001 ***	0.943
	FCU	24.84 ± 6.37 (23.10–26.38)	23.94 ± 6.45 (22.17–25.50)	0.367	1.000	0.108
	FCR	19.32 ± 5.90 (17.67–20.71)	22.86 ± 5.65 (21.35–24.23)	<0.001	0.017 *	0.404
FMM	UT	48.51 ± 7.27 (45.61–51.34)	38.12 ± 8.62 (34.82–41.40)	<0.001	0.002 **	0.509
	AD	35.81 ± 8.82 (32.30–39.14)	44.78 ± 6.57 (42.22–47.34)	<0.001	0.021 *	0.429
	TB	21.21 ± 7.41 (18.42–24.17)	21.29 ± 5.73 (19.09–23.55)	0.780	1.000	0.036
	BB	31.76 ± 8.13 (28.59–34.98)	32.24 ± 5.61 (30.1–34.42)	0.447	1.000	0.098
	ECRL	37.10 ± 8.46 (33.85–40.34)	42.03 ± 6.02 (39.68–44.39)	0.004	0.099	0.370
	PQ	26.52 ± 6.70 (23.95–29.26)	37.25 ± 5.99 (34.89–39.57)	<0.001	<0.001 ***	0.704
	FCU	24.44 ± 8.15 (21.36–27.62)	27.67 ± 6.35 (25.23–30.17)	0.062	1.000	0.241
	FCR	20.37 ± 6.36 (17.98–22.89)	22.54 ± 4.51 (20.83–24.32)	0.066	1.000	0.238
GMFC	UT	47.57 ± 8.06 (42.29–52.56)	38.08 ± 8.62 (32.70–43.59)	0.051	1.000	0.564
	AD	35.83 ± 8.56 (30.19–41.21)	38.76 ± 6.83 (34.40–43.02)	0.377	1.000	0.255
	TB	21.76 ± 6.19 (18.02–25.87)	20.20 ± 5.83 (16.67–24.07)	0.700	1.000	0.111
	BB	31.84 ± 6.14 (27.97–35.79)	38.02 ± 5.79 (34.21–41.52)	0.004	0.093	0.834
	ECRL	36.35 ± 7.73 (31.45–41.22)	39.66 ± 5.04 (36.49–42.86)	0.202	1.000	0.368
	PQ	29.51 ± 6.23 (25.54–33.61)	36.93 ± 4.87 (33.89–40.06)	<0.001	0.031 *	0.930
	FCU	21.45 ± 5.36 (18.10–24.87)	25.61 ± 5.47 (22.26–29.08)	0.133	1.000	0.433
	FCR	18.75 ± 4.97 (15.65–21.91)	20.88 ± 2.60 (19.26–22.56)	0.301	1.000	0.298

Notes: UT, Upper trapezius; AD, Anterior deltoid; TB, Triceps brachii; BB, Biceps brachii; ECRL, Extensor carpi radialis longus; PQ, Pronator quadratus; FCU, Flexor carpi ulnaris; FCR, Flexor carpi radialis. Values are presented as mean ± standard deviation (SD) and 95% confidence interval (CI). p-values were obtained using the Wilcoxon signed-rank test for paired comparisons between PS and NPS. Adjusted p-values were calculated using the Bonferroni correction, with values >1.000 set to 1.000. Effect sizes were computed as $r={\displaystyle \frac{\mathrm{Z}}{\sqrt{N}}},$ where N is the number of paired observations. * p < 0.05, ** p < 0.01, *** p < 0.001 (Bonferroni-adjusted p-values).

presents the mean muscle activation levels (%) of upper limb muscles for the PS and NPS across the three functional domains of the WMFT. Bonferroni-adjusted p-values and effect sizes (r) are also reported. During PRT tasks, the UT exhibited significantly higher activation on the PS compared to the NPS (adj. p < 0.001, r = 0.553), reflecting increased proximal muscle recruitment on the affected side. In contrast, the ECRL showed significantly higher activation on the NPS (adj. p < 0.001, r = 0.878), consistent with preserved or compensatory wrist extensor activity. The PQ also demonstrated significantly greater activation on the NPS (adj. p < 0.001, r = 0.943), suggesting reduced pronator engagement on the affected side. The FCR showed significantly higher activation on the NPS (adj. p = 0.017, r = 0.404), whereas no significant differences were found for the AD, TB, BB, and FCU after Bonferroni correction. In FMM tasks, significant interlimb asymmetries were observed for the UT (adj. p = 0.002, r = 0.509) and AD (adj. p = 0.021, r = 0.429). The UT exhibited greater activation on the PS, likely reflecting compensatory proximal recruitment, whereas the AD showed higher activation on the NPS. The PQ displayed markedly lower activation on the PS (adj. p < 0.001, r = 0.704), indicating reduced pronator engagement. Although the ECRL reached marginal significance before correction (p = 0.004), this effect did not remain significant after adjustment (adj. p = 0.099). No significant differences were observed for the TB, BB, FCU, or FCR. For GMFC tasks, most muscles exhibited symmetrical activation patterns after adjustment. The PQ showed significantly lower activation on the PS compared to the NPS (adj. p = 0.031, r = 0.930), consistent with persistent pronator weakness even during gross functional movements. The BB displayed a marginal difference before correction (p = 0.004), but this did not remain significant after adjustment (adj. p = 0.093). All other muscles, including UT, AD, TB, ECRL, FCU, and FCR, revealed no significant interlimb differences following Bonferroni correction.

3.2. Joint Mobility

Table 3. Differences in joint range of motion (degrees) of the upper limbs between the paretic side (PS) and non-paretic side (NPS) during proximal reaching transport (PRT), fine motor manipulation (FMM), and gross motor functional control (GMFC) tasks of the Wolf Motor Function Test (WMFT), with Bonferroni-adjusted p-values and effect sizes (r).

Range of Motion		PS (95% CI)	NPS (95% CI)	p-Value	Adjusted p (Bonferroni)	r
PRT	SF	32.76 ± 11.80 (27.95–37.78)	36.93 ± 13.08 (31.47–42.73)	0.008	0.203	0.760
	SA	44.19 ± 24.73 (34.85–55.56)	41.94 ± 22.24 (33.64–52.67)	0.510	1.000	0.190
	SER	61.22 ± 38.66 (46.23–79.16)	42.46 ± 13.07 (36.99–47.90)	0.703	1.000	0.110
	EF	34.73 ± 11.92 (29.79–39.99)	44.53 ± 12.68 (39.22–49.95)	0.001	0.025 *	0.946
	WF	24.29 ± 9.37 (20.45–28.45)	25.94 ± 9.16 (22.08–29.95)	0.457	1.000	0.215
	WRD	14.26 ± 4.19 (12.50–16.08)	16.44 ± 5.55 (14.11–18.83)	0.163	1.000	0.402
	WS	77.28 ± 45.66 (59.08–97.21)	47.68 ± 24.85 (38.16–59.22)	0.013	0.301	0.721
	UAP	31.72 ± 11.11 (27.18–36.67)	30.81 ± 8.87 (27.10–34.60)	0.465	1.000	0.211
	UAR	43.88 ± 14.42 (37.77–50.19)	40.78 ± 12.21 (35.49–46.05)	0.666	1.000	0.125
	FER	68.39 ± 38.59 (52.97–86.07)	68.61 ± 41.35 (52.34–87.20)	0.573	1.000	0.163
FMM	SF	52.36 ± 16.52 (44.08–63.60)	52.25 ± 8.32 (47.36–57.28)	0.142	1.000	0.329
	SA	49.98 ± 17.30 (40.27–61.33)	44.50 ± 12.93 (37.20–52.61)	0.378	1.000	0.197
	SER	62.93 ± 41.02 (41.17–90.02)	42.94 ± 9.02 (37.54–48.37)	0.995	1.000	0.001
	EF	39.71 ± 8.00 (35.04–44.49)	55.28 ± 7.34 (51.14–59.82)	<0.001	0.003 **	0.865
	WF	44.19 ± 10.09 (38.35–50.40)	47.52 ± 8.00 (42.81–52.26)	0.371	1.000	0.199
	WRD	30.60 ± 7.13 (26.35–34.89)	31.80 ± 6.86 (27.80–36.02)	0.827	1.000	0.049
	WS	77.83 ± 18.53 (67.17–89.09)	73.40 ± 26.81 (60.09–91.37)	0.138	1.000	0.331
	UAP	41.68 ± 11.13 (35.55–48.76)	38.11 ± 7.36 (33.87–42.66)	0.965	1.000	0.010
	UAR	52.48 ± 11.79 (45.77–60.04)	45.43 ± 8.66 (40.52–50.66)	0.156	1.000	0.317
	FER	98.45 ± 45.38 (73.07–126.90)	106.37 ± 56.32 (74.95–142.60)	0.886	1.000	0.032
GMFC	SF	51.53 ± 7.89 (44.11–58.90)	57.16 ± 10.12 (47.02–66.27)	0.234	1.000	0.420
	SA	59.77 ± 26.59 (38.70–88.08)	71.76 ± 31.81 (43.87–95.80)	0.379	1.000	0.311
	SER	50.08 ± 13.61 (38.11–63.56)	62.14 ± 13.86 (48.55–75.33)	0.134	1.000	0.530
	EF	54.08 ± 9.79 (44.67–63.17)	67.08 ± 10.04 (57.45–76.60)	0.070	1.000	0.640
	WF	46.54 ± 12.29 (35.31–58.47)	56.87 ± 10.38 (46.76–66.64)	0.278	1.000	0.384
	WRD	33.43 ± 6.52 (27.45–39.75)	35.57 ± 6.89 (29.34–42.44)	0.836	1.000	0.073
	WS	70.91 ± 13.49 (58.57–83.78)	84.41 ± 12.28 (72.77–95.79)	0.179	1.000	0.475
	UAP	36.81 ± 6.54 (30.53–43.17)	43.13 ± 7.89 (35.28–50.35)	0.163	1.000	0.494
	UAR	44.37 ± 8.43 (36.51–52.35)	47.90 ± 11.55 (38.76–60.16)	0.836	1.000	0.073
	FER	67.33 ± 19.30 (50.46–87.31)	100.87 ± 49.07 (59.55–152.38)	0.379	1.000	0.311

Notes: SF, Shoulder flexion; SA, Shoulder abduction; SER, Shoulder external rotation; EF, Elbow flexion; WF, Wrist flexion; WRD, Wrist radial deviation; WS, Wrist supination; UAP, Upper arm pitch; UAR, Upper arm roll, FER, Forearm external rotation. Values are presented as mean ± standard deviation (SD) and 95% confidence interval (CI). p-values were obtained using the Wilcoxon signed-rank test for paired comparisons between PS and NPS. Adjusted p-values were calculated using the Bonferroni correction, with values > 1.000 set to 1.000. Effect sizes were computed as $r={\displaystyle \frac{\mathrm{Z}}{\sqrt{N}}},$ where N is the number of paired observations. * p < 0.05, ** p < 0.01 (Bonferroni-adjusted p-values).

summarizes the mean ROM values (degrees) for upper limb segments on the PS and NPS across the three functional domains of the WMFT. Bonferroni-adjusted p-values and effect sizes (r) are also reported. During PRT tasks, only EF remained significant after Bonferroni correction (adj. p = 0.025, r = 0.946), indicating a pronounced reduction in elbow flexion ROM on the PS compared to the NPS. This suggests a marked limitation in elbow flexion mobility during proximal reaching. In contrast, SF (p = 0.008, adj. p = 0.203, r = 0.760) and WS (p = 0.013, adj. p = 0.301, r = 0.721) reached nominal significance before correction but did not remain significant after adjustment. Although these variables showed large effect sizes, they are better interpreted as trends rather than clear asymmetries. All other ROM variables, including SA, SER, WF, WRD, UAP, UAR, and FER, revealed no significant interlimb differences, suggesting largely symmetrical mobility for these joints during proximal reaching. In FMM tasks, EF remained significant after Bonferroni correction (p < 0.001, adj. p = 0.003, r = 0.865), reflecting a substantial reduction in elbow flexion ROM on the PS compared to the NPS. This indicates a consistent limitation in elbow flexion mobility during fine manipulation tasks. All other ROM variables, including SF, SA, SER, WF, WRD, WS, UAP, UAR, and FER, showed no significant differences between sides after adjustment (adj. p = 1.000). Although some variables, such as SF, WS, and UAR, exhibited small-to-moderate effect sizes, none reached statistical significance, suggesting relatively preserved mobility for these joints during fine manipulation tasks. For GMFC tasks, none of the measured ROM variables showed significant interlimb differences after Bonferroni correction (adj. p = 1.000 for all). Several variables, including EF (r = 0.640), SF (r = 0.420), SER (r = 0.530), UAP (r = 0.494), and WS (r = 0.475), showed moderate effect sizes, but none reached statistical significance. This indicates that gross motor functional control was generally performed with symmetrical joint mobility between the PS and NPS. While absolute ROM values tended to be lower on the PS for most joints, the differences were small and not statistically meaningful.

3.3. Movement Amplitude

Table 4. Differences in the root mean square (m/s²) values of the upper limbs between the paretic side (PS) and non-paretic side (NPS) during proximal reaching transport (PRT), fine motor manipulation (FMM), and gross motor functional control (GMFC) tasks of the Wolf Motor Function Test (WMFT), with Bonferroni-adjusted p-values and effect sizes (r).

Root Mean Square		PS	NPS	p-Value	Adjusted p (Bonferroni)	r
PRT	UAX	671.82 ± 108.04 (625.34–718.31)	727.59 ± 102.48 (684.50–771.68)	0.002	0.044 *	0.900
	UAY	283.36 ± 65.78 (255.05–311.66)	229.90 ± 51.58 (207.71–252.09)	0.007	0.177	0.773
	UAZ	624.22 ± 94.62 (583.51–664.94)	588.50 ± 93.36 (548.33–628.67)	0.092	1.000	0.486
	FAX	301.41 ± 84.01 (265.26–337.55)	290.50 ± 78.00 (256.94–324.06)	0.975	1.000	0.009
	FAY	409.01 ± 70.28 (378.77–439.25)	432.65 ± 73.80 (400.90–464.41)	0.398	1.000	0.244
	FAZ	858.92 ± 61.23 (832.58–885.27)	858.26 ± 64.72 (830.41–886.11)	0.771	1.000	0.084
	HX	337.53 ± 85.91 (300.57–374.49)	322.36 ± 95.71 (281.18–363.55)	0.928	1.000	0.026
	HY	310.57 ± 72.71 (279.29–341.86)	287.67 ± 80.70 (252.94–322.39)	0.005	0.117	0.813
	HZ	915.97 ± 46.07 (896.15–935.80)	940.29 ± 66.99 (911.47–969.11)	0.001	0.015 *	0.987
FMM	UAX	700.22 ± 81.79 (650.75–749.70)	779.48 ± 66.94 (738.99–819.97)	0.029	0.699	0.488
	UAY	341.76 ± 64.90 (302.50–381.01)	264.13 ± 62.14 (226.54–301.71)	<0.001	0.006 **	0.815
	UAZ	573.74 ± 94.72 (516.44–631.03)	535.66 ± 68.76 (494.07–577.25)	0.291	1.000	0.236
	FAX	322.18 ± 76.30 (276.03–368.33)	344.07 ± 50.10 (313.77–374.38)	0.237	1.000	0.264
	FAY	457.30 ± 68.97 (415.58–499.02)	513.46 ± 82.57 (463.51–563.41)	0.047	1.000	0.443
	FAZ	814.13 ± 62.37 (776.41–851.86)	787.90 ± 82.65 (737.91–837.89)	0.385	1.000	0.194
	HX	368.15 ± 68.20 (326.89–409.40)	352.50 ± 60.21 (316.08–388.92)	0.472	1.000	0.161
	HY	407.01 ± 53.62 (374.58–439.44)	396.72 ± 78.19 (349.42–444.01)	0.345	1.000	0.211
	HZ	841.13 ± 50.85 (810.38–871.89)	879.06 ± 59.62 (842.99–915.12)	0.027	0.656	0.493
GMFC	UAX	696.46 ± 67.38 (630.43–762.49)	732.98 ± 64.32 (669.94–796.01)	0.098	1.000	0.414
	UAY	323.61 ± 58.26 (266.51–380.71)	288.58 ± 44.78 (244.69–332.46)	0.379	1.000	0.220
	UAZ	600.06 ± 97.44 (504.56–695.55)	588.65 ± 85.70 (504.67–672.63)	0.756	1.000	0.078
	FAX	360.11 ± 73.04 (288.53–431.69)	359.42 ± 71.89 (288.96–429.87)	0.408	1.000	0.207
	FAY	547.87 ± 71.77 (477.53–618.20)	602.25 ± 83.21 (520.71–683.80)	0.234	1.000	0.297
	FAZ	744.33 ± 86.93 (659.14–829.52)	714.08 ± 86.71 (629.11–799.05)	0.605	1.000	0.129
	HX	414.50 ± 75.89 (340.13–488.88)	404.83 ± 67.35 (338.82–470.83)	0.535	1.000	0.401
	HY	526.80 ± 56.76 (471.17–582.42)	499.32 ± 73.70 (427.10–571.55)	0.326	1.000	0.246
	HZ	759.44 ± 74.08 (686.84–832.04)	812.55 ± 58.32 (755.39–869.70)	0.044	1.000	0.504

Notes: UAX, Upper arm X; UAY, Upper arm Y; UAZ, Upper arm Z; FAX, Forearm X; FAY, Forearm Y; FAZ, Forearm Z; HX, Hand X; HY, Hand Y; HZ, Hand Z. Values are presented as mean ± standard deviation (SD) and 95% confidence interval (CI). p-values were obtained using the Wilcoxon signed-rank test for paired comparisons between PS and NPS. Adjusted p-values were calculated using the Bonferroni correction, with values > 1.000 set to 1.000. Effect sizes were computed as $r={\displaystyle \frac{\mathrm{Z}}{\sqrt{N}}},$ where N is the number of paired observations. * p < 0.05, ** p < 0.01 (Bonferroni-adjusted p-values).

presents the RMS values (m/s²) of upper limb segment movements across the X, Y, and Z axes during PRT, FMM, and GMFC tasks, comparing the PS and NPS. During PRT tasks, significant asymmetries were found only for upper arm medio-lateral acceleration (UAX) and hand vertical acceleration (HZ) after Bonferroni correction. UAX was significantly lower on the PS compared to the NPS (adj. p = 0.044, r = 0.900), consistent with restricted medio-lateral excursion of the proximal segment during reaching. Similarly, HZ was significantly lower on the PS (adj. p = 0.015, r = 0.987), reflecting reduced vertical oscillation of the hand during transport. In FMM tasks, a significant interlimb difference was observed only for UAY (adj. p = 0.006, r = 0.815), with greater medio-lateral displacement on the PS. This pattern suggests exaggerated side-to-side excursion of the proximal segment during fine motor manipulation, potentially reflecting compensatory control strategies. No other upper limb segments showed significant interlimb differences after Bonferroni correction. For GMFC tasks, no significant interlimb differences were observed in RMS values for any upper limb segment after Bonferroni correction. This indicates that gross motor functional control tasks were performed with largely symmetrical joint excursions between the PS and NPS.

3.4. Correlation Analysis

3.4.1. Correlations Among EMG Variables

Table 5. Significant Spearman correlations among electromyography (EMG) variables during proximal reaching transport (PRT), fine motor manipulation (FMM), and gross motor functional control (GMFC) tasks of the Wolf Motor Function Test (WMFT) for the paretic side (PS) and non-paretic side (NPS).

Functional Domain		Variable 1	Variable 2	r	p-Value
PRT	PS	AD	FCU	0.685	0.029 *
		ECRL	PQ	0.830	0.003 **
	NPS	BB	PQ	0.721	0.019 *
		ECRL	PQ	0.818	0.004 **
FMM	PS	ECRL	PQ	0.721	0.019 *
			FCR	0.673	0.033 *
		PQ	FCR	0.697	0.025 *
		FCU	FCR	0.758	0.011 *
	NPS	AD	ECRL	−0.636	0.048 *
		PQ	FCU	0.673	0.033 *
			FCR	0.685	0.029 *
		FCU	FCR	0.782	0.008 **
GMFC	PS	UT	BB	0.857	0.007 **
			ECRL	0.762	0.028 *
		BB	PQ	0.810	0.015 *
		ECRL	FCR	0.762	0.028 *
		PQ	FCR	0.905	0.002 **
		FCU	FCR	0.738	0.037 *
	NPS	TB	ECRL	0.929	0.001 **
		BB	ECRL	0.810	0.015 *
			PQ	0.857	0.007 **
		FCU	FCR	0.905	0.002 **

Notes: UT, Upper trapezius; AD, Anterior deltoid; TB, Triceps brachii; BB, Biceps brachii; ECRL, Extensor carpi radialis longus; PQ, Pronator quadratus; FCU, Flexor carpi ulnaris; FCR, Flexor carpi radialis. * p < 0.05, ** p < 0.01.

summarizes significant Spearman’s rank correlations among upper limb muscle activation variables (EMG) during PRT, FMM, and GMFC tasks for both the PS and NPS. Only statistically significant associations are reported. During PRT tasks, the PS demonstrated strong positive correlations between AD–FCU (r = 0.685, p = 0.029) and ECRL–PQ (r = 0.830, p = 0.003). On the NPS, significant correlations were identified between BB–PQ (r = 0.721, p = 0.019) and ECRL–PQ (r = 0.818, p = 0.004), reflecting coordinated activation between elbow flexors and forearm stabilizers during reaching and transport. In FMM tasks, the PS exhibited multiple strong correlations among distal muscles, including ECRL–PQ (r = 0.721, p = 0.019), ECRL–FCR (r = 0.673, p = 0.033), PQ–FCR (r = 0.697, p = 0.025), and FCU–FCR (r = 0.758, p = 0.011). These patterns suggest preserved functional coupling between forearm pronators and wrist flexors during fine motor manipulation. On the NPS, similar associations were found between PQ–FCU (r = 0.673, p = 0.033), PQ–FCR (r = 0.685, p = 0.029), and FCU–FCR (r = 0.782, p = 0.008). A notable negative correlation was observed between AD–ECRL (r = −0.636, p = 0.048), possibly reflecting differential recruitment strategies for proximal and distal control. For GMFC tasks, the PS showed strong correlations between UT–BB (r = 0.857, p = 0.007), BB–PQ (r = 0.810, p = 0.015), and PQ–FCR (r = 0.905, p = 0.002). Additional associations included UT–ECRL (r = 0.762, p = 0.028), ECRL–FCR (r = 0.762, p = 0.028), and FCU–FCR (r = 0.738, p = 0.037), consistent with proximal–distal muscle coordination during gross functional tasks such as lifting and folding. On the NPS, widespread strong correlations were identified among forearm muscles and elbow flexors, including TB–ECRL (r = 0.929, p = 0.001), BB–ECRL (r = 0.810, p = 0.015), BB–PQ (r = 0.857, p = 0.007), and FCU–FCR (r = 0.905, p = 0.002).

3.4.2. Correlations Among ROM Variables

Table 6. Significant Spearman correlations among range of motion (ROM) variables during proximal reaching transport (PRT), fine motor manipulation (FMM), and gross motor functional control (GMFC) tasks of the Wolf Motor Function Test (WMFT) for the paretic side (PS) and non-paretic side (NPS).

Functional Domain		Variable 1	Variable 2	r	p-Value
PRT	PS	SF	WF	0.673	0.033 *
			UAP	0.915	<0.001 ***
			UAR	0.709	0.022 *
		SER	WS	0.758	0.011 *
			UAP	0.758	0.011 *
			UAR	0.697	0.025 *
			FER	0.952	<0.001 ***
		WF	UAP	0.697	0.025 *
			FER	0.697	0.025 *
		WS	UAP	0.636	0.048 *
			UAR	0.745	0.013 *
			FER	0.830	0.003 **
		UAP	UAR	0.830	0.003 **
			FER	0.733	0.016 *
		UAR	FER	0.758	0.011 *
	NPS	SF	SER	0.697	0.025 *
			EF	0.758	0.011 *
			WF	0.830	0.003 **
			UAP	0.867	0.001 **
			UAR	0.952	<0.001 ***
		SRE	EF	0.939	<0.001 ***
			UAP	0.733	0.016 *
			UAR	0.770	0.009 **
		EF	UAP	0.733	0.016 *
			UAR	0.806	0.005 **
		WF	UAP	0.636	0.048 *
			UAR	0.697	0.025 *
		WR	FER	0.733	0.016 *
		WS	UAR	0.685	0.029 *
		UAP	UAR	0.903	<0.001 ***
FMM	PS	SF	WF	0.636	0.048 *
			UAR	0.685	0.029 *
			FER	0.685	0.029 *
		SRE	WS	0.721	0.019 *
			UAP	0.648	0.043 *
		UAP	UAR	0.673	0.033 *
		UAR	FER	0.758	0.011 *
	NPS	SF	SER	0.745	0.013 *
			WF	0.915	<0.001 ***
			UAP	0.794	0.006 **
		SRE	UAP	0.661	0.038 *
		EF	FER	0.830	0.003 **
		WF	UAP	0.758	0.011 *
		UAR	FER	0.770	0.009 **
GMFC	PS	SF	EF	0.857	0.007 **
			UAP	0.762	0.028 *
			UAR	0.810	0.015 *
			FER	0.714	0.047 *
		SAB	SER	0.786	0.021 *
			WS	0.786	0.021 *
		SRE	EF	0.905	0.002 **
			WS	0.952	<0.001 ***
			UAR	0.810	0.015 *
		EF	WF	0.738	0.037 *
			WS	0.905	0.002 **
			UAP	0.714	0.047 *
			UAR	0.905	0.002 **
			FER	0.714	0.047 *
		WF	WS	0.786	0.021 *
		WS	UAR	0.738	0.037 *
		UAP	UAR	0.786	0.021 *
		UAR	FER	0.833	0.010 *
	NPS	SF	SER	0.762	0.028 *
			EF	0.905	0.002 **
			WF	0.857	0.007 **
			UAR	0.738	0.037 *
		SRE	WR	0.810	0.015 *
		EF	WF	0.762	0.028 *
		WF	FER	0.738	0.037 *

Notes: SF, Shoulder flexion; SER, Shoulder external rotation; EF, Elbow flexion; WF, Wrist flexion; WS, Wrist supination; UAP, Upper arm pitch; UAR, Upper arm roll, FER, Forearm external rotation. * p < 0.05, ** p < 0.01, *** p < 0.001.

presents significant Spearman correlations among joint ROM variables during the PRT, FMM, and GMFC tasks for both the PS and NPS. Only statistically significant associations are reported. During PRT tasks, the PS showed multiple strong to very strong correlations among shoulder, wrist, and upper arm variables. Notably, SF correlated strongly with UAP (r = 0.915, p < 0.001) and UAR (r = 0.709, p = 0.022), reflecting consistent coupling of gross shoulder motion with upper arm orientation. Similarly, SER was strongly associated with WS and FER, suggesting synergistic control of rotational movement across proximal and distal joints. The NPS also displayed robust correlations, with SF, EF, and SER all linked to UAP, UAR, and FER, consistent with well-integrated multi-joint mobility during reaching. In FMM tasks, the PS revealed significant associations among shoulder movements, wrist flexion, and upper arm orientation. For instance, SF and WF were strongly correlated (r = 0.673, p = 0.033), indicating coordinated flexion across the shoulder and wrist during fine manipulation. The NPS exhibited even stronger associations, with SF and WF showing a very strong correlation (r = 0.915, p < 0.001) and EF tightly linked to FER (r = 0.758, p = 0.011). These findings point to stable segmental coordination during task performance. For GMFC tasks, the PS showed widespread strong correlations among SF, EF, UAP, and FER, highlighting interdependent control of shoulder–elbow–forearm axes during gross functional movements. For example, SF correlated significantly with EF (r = 0.905, p = 0.002), UAP (r = 0.714, p = 0.047) and FER (r = 0.714, p = 0.047), reflecting integrated proximal joint function. On the NPS, significant associations were observed primarily between wrist movements, UAR, and FER, such as WF–FER (r = 0.738, p = 0.037), suggesting coordinated distal segment movement in gross motor tasks.

3.4.3. Correlations Among RMS Variables

Table 7. Significant Spearman correlations among root mean square (RMS) variables during proximal reaching transport (PRT), fine motor manipulation (FMM), and gross motor functional control (GMFC) tasks of the Wolf Motor Function Test (WMFT) for the paretic side (PS) and non-paretic side (NPS).

Functional Domain		Variable 1	Variable 2	r	p-Value
PRT	PS	UAX	UAY	−0.939	<0.001 ***
		FAY	FAZ	−0.842	0.002 **
	NPS	UAX	UAY	−0.939	<0.001 ***
			FAY	−0.782	0.008 **
			HY	−0.673	0.033 *
		UAZ	FAY	0.830	0.003 **
			HY	0.733	0.016 *
		FAX	HX	0.745	0.013 *
		FAY	HY	0.818	0.004 **
		FAZ	HZ	0.952	<0.001 ***
FMM	PS	UAX	UAY	−0.855	0.002 **
		FAX	FAZ	−0.636	0.048 *
		FAY	FAZ	−0.661	0.038 *
			HY	0.782	0.008 **
		HX	HZ	−0.721	0.019 *
	NPS	UAX	UAZ	−0.782	0.008 **
			FAZ	0.648	0.043 *
		UAZ	HX	0.648	0.043 *
		FAY	FAZ	−0.915	<0.001 ***
			HX	0.733	0.016 *
			HY	0.939	<0.001 ***
			HZ	−0.709	0.022 *
		FAZ	HX	−0.685	0.029 *
			HY	−0.879	0.001 **
			HZ	0.842	0.002 *
		HX	HY	0.648	0.043 *
			HZ	−0.733	0.016 *
		HY	HZ	−0.794	0.006 **
GMFC	PS	UAX	UAZ	−0.928	<0.001 ***
		FAX	HX	0.714	0.047 *
		FAY	FAZ	−0.905	0.002 **
	NPS	UAX	UAZ	−0.976	<0.001 ***
			FAX	0.905	0.002 **
		UAZ	FAX	−0.857	0.007 **
		FAX	HX	0.762	0.028 *
		FAY	FAZ	−0.738	0.037 *
			HY	0.881	0.004 **
		FAZ	HZ	0.952	<0.001 ***

Notes: UAX, Upper arm X; UAY, Upper arm Y; UAZ, Upper arm Z; FAX, Forearm X; FAY, Forearm Y; FAZ, Forearm Z; HX, Hand X; HY, Hand Y; HZ, Hand Z. * p < 0.05, ** p < 0.01, *** p < 0.001.

summarizes significant Spearman correlations among RMS values across the X, Y, and Z axes for the upper arm, forearm, and hand segments during the PRT, FMM, and GMFC tasks on both the PS and NPS. Only statistically significant associations are reported. During PRT tasks, the PS showed robust inverse correlations between UAX–UAY (r = −0.939, p < 0.001) and FAY–FAZ (r = −0.842, p = 0.002), indicating compensatory trade-offs between medio-lateral and vertical displacements during reaching. This pattern indicates that increased lateral motion was accompanied by reduced vertical displacements, possibly due to altered motor strategies. On the NPS, similar negative associations were observed between UAX–UAY, UAX–FAY, and UAX–HY, whereas positive correlations emerged between UAZ–FAY, UAZ–HY, FAX–HX, FAY–HY, and FAZ–HZ. These results are consistent with integrated movement across limb segments and axes during coordinated transport. In FMM tasks, the PS exhibited strong negative correlations among upper arm and hand segment axes (e.g., UAX–UAY, HX–HZ), while hand axes such as HX–HY were positively correlated. This suggests directional specificity in distal motor control. On the NPS, very strong negative correlations were found between UAZ–FAZ and FAZ–HY, whereas FAY–HY and FAZ–HZ displayed strong positive associations, reflecting consistent coupling of distal segment during fine motor manipulation. For GMFC tasks, the PS showed strong negative correlations between UAX–UAZ (r = −0.928, p < 0.001) and FAY–FAZ (r = −0.905, p < 0.001), along with a significant positive correlation between FAX–HX. These findings indicate constrained multi-directional movement patterns during gross motor performance, with proximal–distal linkage in anterior–posterior movements. On the NPS, positive correlations were consistently observed between UAX–UAZ, UAZ–FAX, and FAZ–HZ, reflecting coordinated multi-segment movement patterns and symmetrical motor strategies during gross functional control.

3.4.4. Correlations Among EMG, ROM, and RMS Variables

Table 8. Significant Spearman correlations among electromyography (EMG), range of motion (ROM), and root mean square (RMS) variables during proximal reaching transport (PRT), fine motor manipulation (FMM), and gross motor functional control (GMFC) tasks of the Wolf Motor Function Test (WMFT) for the paretic side (PS) and non-paretic side (NPS).

Functional Domain			Variable 1	Variable 2	r	p-Value
PRT	PS	EMG–ROM	UT	WR	−0.673	0.033 *
			AD	SAB	0.770	0.009 **
				WR	−0.770	0.009 **
			FCU	SAB	0.794	0.006 **
			FCR	SAB	0.818	0.004 **
		EMG–RMS	AD	HZ	0.673	0.033 *
			FCU	HZ	0.697	0.025 *
			FCR	HZ	0.721	0.019 *
		ROM–RMS	SAB	HZ	0.745	0.013 *
			EF	HY	0.794	0.006 **
	NPS	EMG–ROM	UT	WR	0.648	0.043 *
			BB	WF	0.709	0.022 *
		EMG–RMS	BB	HZ	0.636	0.048 *
			ECRL	FAY	−0.697	0.025 *
			PQ	HY	−0.636	0.048 *
		ROM–RMS	SF	FAX	0.648	0.043 *
			SAB	UAZ	0.721	0.019 *
				HY	0.648	0.043 *
			WF	FAX	0.636	0.048 *
				HZ	0.685	0.029 *
			WS	FAX	0.648	0.043 *
				HZ	0.673	0.033 *
			FRE	UAY	−0.721	0.019 *
				HX	−0.685	0.029 *
FMM	PS	EMG–ROM	UT	UAR	0.733	0.016 *
				FRE	0.697	0.025 *
			AD	SF	0.721	0.019 *
				FRE	0.818	0.004 **
			ECRL	EF	0.794	0.006 **
			PQ	EF	0.745	0.013 *
			FCR	EF	0.709	0.022 *
		EMG–RMS	AD	FAX	0.636	0.048 *
			BB	FAX	0.685	0.029 *
			ECRL	UAY	−0.818	0.004 **
			FCU	FAX	0.685	0.029 *
		ROM–RMS	EF	HZ	0.685	0.029 *
			WS	UAZ	0.709	0.022 *
			UAP	FAY	−0.697	0.025 *
			FRE	HY	−0.661	0.038 *
	NPS	EMG–ROM	UT	WR	0.794	0.006 **
			BB	SRE	0.636	0.048 *
				UAR	0.636	0.048 *
		EMG–RMS	UT	UAX	0.685	0.029 *
			AD	HX	−0.648	0.043 *
				HZ	0.697	0.025 *
			TB	UAX	0.770	0.009 **
			ECRL	HX	0.721	0.019 *
			PQ	HX	0.648	0.043 *
		ROM–RMS	SRE	UAY	−0.721	0.019 *
			SAB	UAZ	0.782	0.008 **
			WS	FAY	0.818	0.004 **
				FAZ	−0.685	0.029 *
				HY	0.745	0.013 *
			FRE	HX	0.636	0.048 *
GMFC	PS	EMG–ROM	UT	WF	0.762	0.028 *
			AD	WF	0.714	0.047 *
			TB	SAB	−0.810	0.015 *
				WF	−0.881	0.004 **
				WS	−0.762	0.028 *
			BB	SF	0.857	0.007 **
				EF	0.762	0.028 *
				WF	0.762	0.028 *
			PQ	SF	0.857	0.007 **
				SRE	0.714	0.047 *
				EF	0.905	0.002 **
				UAR	0.810	0.015 *
			FCU	SRE	0.714	0.047 *
				WR	0.881	0.004 **
				FRE	0.714	0.047 *
			FCR	SF	0.762	0.028 *
				EF	0.857	0.007 **
				WR	0.810	0.015 *
				UAR	0.714	0.047 *
		ROM–RMS	SAB	HZ	0.738	0.037 *
	NPS	EMG–ROM	BB	FRE	0.929	0.001 **
			ECRL	FRE	0.810	0.015 *
		EMG–RMS	AD	FAZ	0.738	0.037 *
			TB	UAY	−0.738	0.037 *
		ROM–RMS	SF	FAZ	0.714	0.047 *
				HZ	0.738	0.037 *
			SAB	HZ	0.738	0.037 *
			SRE	FAY	−0.786	0.021 *
				FAZ	0.833	0.010 *
				HY	−0.714	0.047 *
				HZ	0.714	0.047 *
			WF	FAZ	0.738	0.037 *
				HZ	0.738	0.037 *
			WS	HY	−0.881	0.004 **
			UAP	HY	−0.738	0.037 *
			UAR	HX	0.786	0.021 *

Notes: UT, Upper trapezius; AD, Anterior deltoid; TB, Triceps brachii; BB, Biceps brachii; ECRL, Extensor carpi radialis longus; PQ, Pronator quadratus; FCU, Flexor carpi ulnaris; FCR, Flexor carpi radialis; SF, Shoulder flexion; EF, Elbow flexion; WF, Wrist flexion; WS, Wrist supination; UAP, Upper arm pitch; UAR, Upper arm roll; UAX, Upper arm X; UAY, Upper arm Y; UAZ, Upper arm Z; FAX, Forearm X; FAY, Forearm Y; FAZ, Forearm Z; HX, Hand X; HY, Hand Y; HZ, Hand Z. * p < 0.05, ** p < 0.01.

summarizes significant Spearman correlations among EMG, ROM, and RMS variables during the PRT, FMM, and GMFC tasks for both the PS and NPS. Only statistically significant associations are reported. During PRT tasks, the PS displayed notable EMG–ROM correlations. UT and AD were negatively associated with WR, whereas AD, FCU, and FCR were positively correlated with SAB. Interestingly, FCR also showed a negative association with SAB, consistent with compensatory muscular patterns despite increased joint excursion. In the EMG–RMS domain, AD, FCU, and FCR correlated positively with vertical hand displacement (HZ). For ROM–RMS relationships, SAB was positively associated with HZ, while EF was strongly correlated with anterior–posterior hand movement (HY). On the NPS, UT showed a positive correlation with WR and BB with WF. In the EMG–RMS domain, BB was positively associated with HZ, while ECRL and PQ were negatively correlated with FAY and HY, respectively. Notably, SAB correlated positively with both UAZ and HY, indicating that shoulder abduction contributed to upward and forward limb motion. In contrast, FRE showed negative correlations with UAY and HX, reflecting altered rotational dynamics during transport. In FMM tasks, the PS revealed several cross-domain associations. In the EMG–ROM domain, UT and AD showed positive correlations with UAR and FRE, whereas ECRL, PQ, and FCR exhibited strong associations with EF. For EMG–RMS, AD and BB were correlated positively with FAX, whereas ECRL was negatively associated with UAY. FCU also showed a positive association with FAX, and EF correlated with HZ. ROM–RMS correlations revealed a positive association between WS and UAZ. In contrast, UAP and FRE showed negative correlations with FAY and HY. This pattern points to distinct segmental contributions to fine motor control. On the NPS, UT and BB showed positive associations with SRE and UAR, respectively. In the EMG–RMS domain, UT, AD, and TB showed positive correlations with UAX and HZ, whereas ECRL and PQ were linked to HX. Strong negative correlations were observed between SRE–UAY and SAB–UAZ. ROM–RMS correlations revealed that WS showed a positive association with FAZ, while FRE exhibited positive correlations with both HX and HY. For GMFC tasks, the PS exhibited consistent EMG–ROM correlations, with UT, AD, TB, and BB displaying strong associations with WF. SF exhibited strong correlations with both EF and WF. Additionally, PQ, FCU, and FCR were positively correlated with SRE, EF, WR, and UAR. This finding underscores a close relationship between joint motion and muscular effort. In the EMG–RMS domain, BB and ECRL showed very strong positive associations with FRE. In terms of ROM–RMS associations, SAB showed a positive correlation with HZ. On the NPS, AD and TB showed correlations with FAZ and UAY, respectively. This suggests proximal–distal movement coupling. ROM–RMS relationships identified strong associations between SF, SAB, and FAZ. Additional correlations were observed between SRE and FAZ (positive) and between SRE and HY (negative). WS showed a strong negative correlation with HY, while UAR exhibited a positive correlation with HX.

4. Discussion

This study examined the value of categorizing WMFT tasks into three functional domains: proximal reaching and transport (PRT), fine motor manipulation (FMM), and gross motor functional control (GMFC). This categorization highlights the different neuromuscular and kinematic demands on the paretic upper limb after stroke. Each domain emphasizes specific aspects of motor performance. Examples include shoulder-based stabilization and transport, precise distal control, and large-amplitude gross movements. We analyzed these domains separately and reported only results that remained significant after correction for multiple comparisons. This domain-based approach revealed distinct motor adaptations and offered new insight into post-stroke upper limb function.

There are several limitations to consider when interpreting these findings. First, the sample size was relatively small. All participants were in the chronic phase of stroke. Their mean disease duration was 13.7 years, and they showed mild-to-moderate upper limb impairment. Because of this long disease duration, the observed neuromuscular and kinematic patterns may reflect well-established compensatory strategies rather than early recovery mechanisms. As a result, caution is needed when generalizing to individuals with more severe deficits or those in the acute or subacute stages. Second, the cross-sectional design limits any causal interpretation of the relationships among muscle activation, joint mobility, and movement amplitude. Third, categorizing WMFT tasks into PRT, FMM, and GMFC domains enabled domain-specific analysis. However, this method may oversimplify the continuous and overlapping demands of everyday movements. Future work should consider representing task complexity as a continuum. Lastly, the use of sensor-derived variables (EMG, ROM, RMS) provided detailed quantitative insights. However, the absence of complementary measures, such as trunk kinematics, clinical scores, and patient-reported outcomes, restricts the interpretation of these findings in terms of real-world function and quality of life.

Given these boundaries, this study should be considered as a preliminary proof-of-concept investigation. It demonstrates the feasibility and utility of integrating sensor-based measures into WMFT task analysis. Future research should include additional measures and adopt more naturalistic designs to better inform individualized rehabilitation strategies.

PRT performance was marked by an over-reliance on proximal stabilizers. Specifically, UT activation was greater on the PS than on the NPS. This compensatory shoulder elevation likely supports goal-directed reaching despite reduced distal motor control, consistent with prior observations in stroke rehabilitation [36,37]. In contrast, distal muscles such as the ECRL and FCR showed reduced activation, reflecting the vulnerability of fine motor effectors to corticospinal tract damage [38]. Kinematic analysis revealed reduced shoulder and elbow flexion ROM on the PS, which constrained limb elevation and forward reach. Such restrictions have been linked to difficulties in achieving full reaching postures [39]. RMS analysis further showed diminished medio-lateral displacement of the upper arm, indicating low-variability and stereotyped trajectories [40]. The negative correlation between UT activation and wrist displacement amplitude supports the well-documented proximal–distal coupling in post-stroke reaching [41]. Collectively, these results describe a movement strategy characterized by proximal overactivation, distal underuse, limited joint mobility, and simplified trajectory control. Clinically, this highlights the importance of interventions that target proximal–distal coordination and encourage flexible, multidirectional reaching. These findings directly support H1, which predicted greater proximal–distal asymmetries in PRT tasks compared with GMFC.

FMM tasks posed unique challenges, with several deficits persisting even after statistical correction. On the PS, UT activation remained elevated during fine manipulation. This indicates that proximal compensatory strategies continue to operate even in precision-oriented contexts [42,43]. At the same time, distal and mid-level muscles such as the ECRL and PQ showed reduced activation, reflecting their selective vulnerability to corticospinal damage [32]. The PS also displayed diminished elbow flexion ROM, which limits the ability to position the hand within the functional workspace. These impairments may foster compensatory trunk or shoulder strategies and reinforce maladaptive postures over time [44]. Amplitude reductions were observed in both medio-lateral upper arm motion and vertical hand displacement, suggesting restricted adaptability across directions. Higher UT activation was also negatively associated with vertical hand displacement, highlighting a proximal–distal trade-off consistent with prior sensor-based findings [45]. Overall, these results show that post-stroke fine motor performance is marked by excessive proximal reliance, weakened distal engagement, reduced elbow mobility, and constrained movement variability. Clinically, this underscores the need for rehabilitation programs that integrate proximal–distal coordination and promote adaptive multidirectional control. In addition, the results support H1 by demonstrating pronounced asymmetries. They also support H3 by showing strong correlations between proximal overuse and reduced distal movement amplitude.

GMFC tasks were largely characterized by symmetrical muscle activation patterns between limbs. This suggests that gross transport movements place fewer demands on compensatory proximal activation than either PRT or FMM. Nevertheless, selective deficits were evident. On the PS, activation of the BB and PQ was reduced, indicating residual impairments in elbow flexion and forearm pronation [46,47]. Joint ROM measures did not reveal significant asymmetries, suggesting that gross mobility was preserved. However, reduced vertical displacement of the hand on the PS pointed to subtle lifting and distal coordination deficits. These deficits may remain hidden during low-complexity tasks but still limit performance when fine adjustments are required [48]. These results align with prior evidence that gross transport capacity is often maintained despite persistent distal control deficits [49,50]. From a clinical perspective, this underscores the importance of maintaining distal motor training. Such training is necessary even in patients who show apparently preserved gross reaching function [51,52]. These observations support H2, which predicted overall symmetry in GMFC with residual distal motor deficits.

In addition to observational assessment, this study incorporated EMG, ROM, and RMS measures. This approach offered a multidimensional view of post-stroke upper limb performance [53,54]. The conventional WMFT relies on ordinal scoring and completion time. In contrast, our sensor-based approach provides objective quantification of muscle activation, joint kinematics, and movement amplitude. This facilitates the detection of subtle neuromuscular and kinematic deficits, such as proximal–distal asymmetries, restricted joint excursions, and reduced movement variability. These features may not be evident through visual assessment alone [55]. Moreover, categorizing tasks into PRT, FMM, and GMFC supports task-specific profiling of motor performance. This allows domain-level impairments to be identified even when total WMFT scores appear normal. The observed coupling between proximal overuse and distal underuse is noteworthy. It reinforces the importance of designing individualized interventions that address both ends of the limb in a coordinated manner [54]. Objective sensor metrics also offer the potential for sensitive progress monitoring over time. They can detect changes in neuromuscular coordination before such changes are reflected in WMFT scores. Furthermore, detailed sensor-derived profiling provides a foundation for developing task-specific training protocols. It also supports the design of adaptive biofeedback systems tailored to each patient’s movement profile. Ultimately, leveraging real-time sensor analytics may enhance motor relearning and foster flexible and efficient movement strategies. In turn, these improvements can support greater functional independence in daily activities [56]. These integrated findings reinforce H3. Significant EMG–ROM–RMS correlations confirm the interdependence of proximal and distal impairments.

Because the mean disease duration in our cohort was long, the observed neuromuscular and kinematic patterns probably represent established compensatory strategies rather than early-recovery. These domain-specific impairments are relevant for chronic stroke rehabilitation. However, caution is needed when generalizing to subacute or acute populations.

5. Conclusions

This study quantitatively demonstrated the value of categorizing the Wolf Motor Function Test (WMFT) into three functional domains: proximal reaching and transport (PRT), fine motor manipulation (FMM), and gross motor functional control (GMFC). This framework revealed domain-specific neuromuscular and kinematic features of upper limb performance after stroke. Real-time measurements of electromyography (EMG), range of motion (ROM), and root mean square (RMS) acceleration were integrated into the analysis. Together, these measures provided a multidimensional view of how compensatory strategies manifest across different task types.

Across domains, stroke patients exhibited reduced distal muscle activation (20–30% lower %RVC compared to the non-paretic side) and limited joint excursion (15–25° reduction in ROM). In contrast, proximal stabilizers such as the upper trapezius and biceps demonstrated over-recruitment (up to 25%RVC) to compensate for impaired distal control. In PRT and FMM tasks, asymmetries were most pronounced. Moderate-to-strong correlations were observed across modalities, such as EMG–ROM (e.g., FCR–SAB: r = 0.818, p = 0.004) and EMG–RMS (e.g., FCU–HZ: r = 0.697, p = 0.025). These results confirm the interplay between excessive proximal activation and restricted distal mobility. In contrast, GMFC tasks elicited more symmetrical muscle activity and greater joint excursions. These patterns were supported by very high EMG–ROM associations (e.g., TB–WF: r = −0.881, p = 0.004; PQ–EF: r = 0.905, p = 0.002). However, subtle deficits in fine motor control still persisted. Notably, several inter-modality correlations exceeded r = 0.80, peaking at r = 0.929 (p = 0.001). These findings underscore the tight coupling of neuromuscular and kinematic mechanisms in post-stroke motor coordination.

• Practical implications: This sensor-based, domain-specific approach quantifies both absolute impairments (e.g., 20–30% lower EMG activity, 15–25° ROM reduction) and inter-modality associations. As a result, it enables precise detection of deficits that conventional WMFT scoring may overlook. It further supports individualized rehabilitation strategies that address proximal compensation and distal coordination together.
• Scientific implications: Integrating EMG, ROM, and RMS measures offers a robust, multidimensional framework for quantifying motor coordination after stroke. This approach also reveals mechanisms underlying compensatory strategies. Together, these insights advance understandings of motor control deficits and guide the development of evidence-based rehabilitation protocols.
• Societal implications: This approach allows earlier detection and more precise targeting of persistent impairments, even long after stroke onset. These benefits may improve long-term functional independence. They may also help reduce disability and ease the societal and economic burden of chronic stroke.

Future research should include a broader spectrum of participants, covering acute, subacute, and more severely impaired individuals. It should also use ecological assessments in real-world contexts. In addition, patient-reported outcomes should be combined with complementary modalities, such as trunk kinematics and kinetic analysis. Such steps will improve the applicability, reliability, and translational potential of the sensor-based WMFT framework.