2.1 Participants
The study cohort consisted of 21 stroke patients and 21 healthy adults. For the stroke group, the inclusion criteria were to: i) be aged between 18 and 90 years old, ii) be at more than 3 months of a first cerebrovascular accident of any aetiology (hemorrhagic or ischemic), and iii) have an UL motor impairment with FM-UE ≥ 15 (36). The non-inclusion criteria were to: i) have hemineglect or severe attentional problems (omission of more than 15 bells on the Bell’s test; (37), ii) have aphasia of comprehension dysfunction (Boston Diagnostic Aphasia Examination < 4/5; (38), and iii) have severe cognitive dysfunction (Mini Mental State Examination-MMSE < 24; (39). To be included, the healthy adults had to be aged between 60 and 90 years old (to fit with the stroke group age) and to be right-handed assessed by the Edinburgh Handedness Inventory (40). Exclusion criteria were the existence of neurological (including a history of traumatic brain injury) or motor disorders at the level of the upper limb (history of tendinous disease, arthritis, surgery). Healthy participants were recruited via local association, while stroke ones were recruited at the beginning of a rehabilitation protocol (ReArm project, Clinical trial identifier: NCT04291573, 2nd March 2020).
Table 1 provides detailed participant information, including gender, age, lesioned side, laterality, and clinical scores (refer to the clinical assessments section for additional details). For the stroke group, Table 2 presents all patients' demographic data and clinical history.
In accordance with the Declaration of Helsinki, this study was approved from the French Research Ethics Committee, (Comité de Protection des Personnes-CPP SUD-EST II, N°ID-RCB: 2019-A00506-51, http://www.cppsudest2.fr/) for the stroke patients, and from the local Ethics Committee of the EuroMov DHM laboratory for the healthy subjects (EuroMov IRB, number 1912B). All participants provided informed written consent prior participation in the study.
Table 1. Characteristics of the participants for each group (n=21)
Characteristics
|
Healthy group
|
Stroke group
|
Age (years) (SD)
|
73.1 (± 6.7)
|
64.4 (±10.2)
|
Sex (female/male)
|
11/10
|
6/15
|
Handedness score (SD)
|
0.96 (± 0.08)
|
-
|
Paretic arm (right/left)
|
-
|
8/13
|
FM-UE
|
-
|
48.7 (±5.9)
|
WMFT
|
-
|
57.3 (±9.8)
|
BBT ratio
|
-
|
54.0 (±25.1)
|
BBT ratio = (paretic score / non-paretic score) * 100
Table 2. Demographic information, clinical data, lesion information and clinical scores.
P
|
Age
|
Gender
|
Hemisphere lesioned
|
HD before stroke
|
Paretic arm
|
FM-UE
|
BI
|
Type of stroke
|
1
|
62
|
M
|
L
|
R
|
R
|
45
|
85
|
Is
|
2
|
61
|
M
|
R
|
R
|
L
|
55
|
95
|
Is
|
3
|
52
|
M
|
R
|
R
|
L
|
51
|
90
|
Is
|
4
|
63
|
M
|
L
|
L
|
R
|
44
|
95
|
Is
|
5
|
70
|
M
|
R
|
R
|
L
|
51
|
100
|
Is
|
6
|
73
|
F
|
R
|
R
|
L
|
53
|
-
|
H
|
7
|
63
|
F
|
R
|
R
|
L
|
27
|
90
|
Is
|
8
|
57
|
F
|
R
|
R
|
L
|
60
|
90
|
H
|
9
|
74
|
M
|
R
|
R
|
L
|
50
|
85
|
Is
|
10
|
37
|
M
|
R
|
L
|
L
|
46
|
95
|
Is
|
11
|
68
|
M
|
R
|
R
|
L
|
47
|
95
|
Is
|
12
|
76
|
M
|
L
|
R
|
R
|
41
|
90
|
H
|
13
|
62
|
F
|
R
|
R
|
L
|
45
|
85
|
Is
|
14
|
49
|
F
|
R
|
R
|
L
|
54
|
95
|
Is
|
15
|
82
|
M
|
L
|
R
|
R
|
58
|
100
|
Is
|
16
|
72
|
M
|
L
|
R
|
R
|
44
|
90
|
Is
|
17
|
66
|
M
|
L
|
R
|
R
|
38
|
95
|
Is
|
18
|
73
|
M
|
L
|
R
|
R
|
36
|
25
|
Is
|
19
|
71
|
F
|
L
|
R
|
R
|
57
|
95
|
H
|
20
|
62
|
M
|
R
|
R
|
L
|
46
|
-
|
H
|
21
|
60
|
M
|
R
|
R
|
L
|
43
|
90
|
Is
|
Abbreviations: M, male; F, female; R, right; L, left; HD, hand-dominance; BI, Barthel index (score/100); FM, Upper Limb Fugl-Meyer (score/66); Is, Ischemic; H, Hemorrhagic. The severity of the motor impairment was evaluating using the FM-UE in accordance with the motor impairment classification in clinical and research settings (41).
2.1 Experimental design
Each participant engaged in an hour-long session in a quiet isolated room. The participants were equipped with the fNIRS-fEEG neuroimaging systems and performed two functional UL tasks while seated: a paced reaching arm task and a circular steering task. The setup permitted synchronized recording of UL kinematics and brain activity (fNIRS and fEEG) using lab streaming layer (LSL, https://github.com/labstreaminglayer/App-LabRecorder). More comprehensive details about the functional motor task methodology can be found in our recent methodological paper (see Figure 5 in 42).
2.2 Upper-limb function
All participants performed the two functional UL tasks, as detailed in earlier studies (24,42).
2.2.1 Paced reaching task
Participants were seated on a chair fitted with armrests and were instructed to reach a target (a table tennis ball) placed in front of them at a height of 80 cm and a distance which facilitated the complete extension of the arm. A Kinect sensor (V2, Microsoft, USA), sampled at 30 Hz, was positioned 1.70 m above and 1.60 m away from the target. Participants performed five movements per 20-second block, timed to vocal prompts ("go"; "stop"). After a familiarization block with each arm, participants completed three blocks using their non-dominant/paretic hand, followed by three blocks using their dominant/non-paretic hand. Each block was interspersed with 20 seconds of rest. Then, participants repeated the task for three blocks with each hand under a movement-constrained condition, wherein their shoulders were immobilized to minimize trunk movements.
2.2.2 Circular steering task
This task was based on the speed-accuracy trade-off (43). Participants were seated on a chair in front of a horizontal graphic tablet (A3 size; Wacom, Kazo, Japan) equipped with a stylus affixed to a mouse pad, facing a 24-inch vertical screen projecting a circular target (33-inch circumference) with a 2 cm tunnel. A Kinect was placed above the graphic tablet at the height of 1.70 m. The task was delivered using a lab-made software, the LSL-Mouse (https://github.com/KarimaBak/LSL-Mouse). Participants were instructed to move a cursor as fast as possible in a clockwise direction. During the familiarization phase, participants were instructed to accelerate if errors (any instances outside the 2 cm circular tunnel boundaries) were below 15% (based on pilot testing). The task comprised three blocks for each arm (20 seconds of task with 20 seconds of rest), commencing with their non-dominant/paretic hand.
2.2.3 Clinical assessments of paretic upper limb impairment
In conjunction with the functional kinematics and brain evaluation, patients' UL motor function was appraised through clinical evaluations. We utilized several recognized and validated tests, including the FM-UE (36,41), the Box and Block test (BBT; 44), the Wolf-motor function test (WMFT, 45), the Barthel Index (BI, 46), and the Proximal-arm non-use test (PANU, 3,4). Comprehensive details of these evaluations are described in the cited references.
The FM-UE assesses upper limb motor impairment, while the BBT measures arm and hand grasping function. WMFT evaluates upper limb function, and the BI measures overall functional recovery (independent function in activities of daily living). The PANU test measures the degree to which the paretic upper limb, specifically movements at the shoulder and elbow, is not spontaneously used. These tests collectively provide a comprehensive overview of the paretic UL's functional capacity and impairment (for the FM-UE) level in stroke patients.
Brain activity (fNIRS and fEEG)
Participants wore a custom neoprene head cap equipped with a combined fEEG-fNIRS system to monitor brain activity within the left and right sensorimotor cortical regions during both functional motor tasks. We utilized a wireless Starstim fNIRS integration system (Starstim8, Neuroelectrics, Barcelona, Spain; Octamon+, Artinis Medical Systems, Elst, The Netherlands) to measure fEEG and fNIRS signals. Details regarding the placement of the 16 channels, comprising four fNIRS and four fEEG channels per SM1 hemisphere, are outlined in a previous article (see Figure 1 in 24).
The fEEG electrodes were positioned in and around SM1 cortices: C4, FC2, FC6, CP2 in the right hemisphere and C3, FC1, FC3, CP1 in the left hemisphere, in alignment with the international 10–10 system. The electrodes (NG Geltrode, Neuroelectrics, Spain) were filled with electro-gel (Signa Gel®). Using an ear clip, reference electrodes (CMS, DRL) were placed over the right earlobe. The fEEG signals were sampled at a rate of 500 Hz. We controlled the wifi- fEEG device via a software interface (Neuroelectrics Instrument Controller, NIC v 2.0).
For the fNIRS recording, we used a continuous-wave system employing two wavelengths to capture changes in HbO2 and HbR overlying the left and right SM1, sampling at 10 Hz. The two receivers were positioned at the C1 and C2 locations of the 10–10 fEEG system, with four transmitters placed 3 cm from the receivers using plastic holders. The fNIRS Bluetooth device was managed through a software interface (Oxysoft, v3.2.51.4, Artinis Medical Systems, Elst, The Netherlands).
Following the equipment setup, participants were asked to perform a wrist extension task to verify if the movement induced a hemodynamic response.
2.4 Data analysis
2.4.1 Task performance
The paced reaching and circular steering task kinematics analysis was done based on previous work (3,4,47) and LSL-Kinect software (LSL-KinectV2: https://github.com/KarimaBak/LSL-KinectV2). For the paced reaching task, we calculated the proximal-arm non-use (%) and the hand mean velocity (mm/s). For both tasks, we calculated as trunk compensation parameter, the range of trunk anterior flexion (°) representing the use of the trunk to realize the reaching movement. And, we calculated, as arm use parameters the range of elbow extension (°) representing the use of whole arm to perform the movement.
2.4.2 Brain activity (fNIRS and fEEG)
We processed all fNIRS raw data using the HOMER toolbox in MATLAB (Homer2 NIRS processing package, 50) with the files generated by the Lab Recorder (xdf files). Pre- and post-processing steps are detailed in a previous study (24). We used the relative changes (Δ) in peak HbO2 concentration as an indicator of brain activity.
We analysed all fEEG data using the EEGLAB toolbox on MATLAB (51, version 2021.1), with the files generated by the Lab Recorder (xdf files). Details of pre- and post-processing steps are provided in a previous study (24). We calculated the event-related spectral perturbations (ERSP) in the alpha (8-13 Hz) and beta (14-29 Hz) rhythms, revealing average power changes in these specific time frequencies. This information provides insight into event-related desynchronization (ERD; power decrease in a specific frequency band relative to baseline, i.e., rest) and synchronization (ERS; power increase in a specific frequency band relative to the task). For fEEG and fNIRS analyses, parameters were averaged by tasks (paced reaching; circular steering), hand condition (dominant / non-paretic; non-dominant / paretic), and hemisphere (contralateral / ipsilesional ; ipsilateral / contralesional).
2.5 Statistical analyses
Statistical analyses were performed using R software (version 4.2.1) and the ggplot2 (52), dplyr (53) and rstatix (54) packages. Parametric tests were employed following the validation of data normality via the Shapiro-Wilk test and visual examination of Q-Q plots. Effects sizes were indicated using the partial eta square (η²p), with small (0.02), medium (0.13), and large (0.26) effect sizes noted (55,56) . A threshold of p < .05 was used for statistical significance. If necessary, pairwise comparisons were conducted using t-tests, with the Benjamini-Hochberg procedure applied for p-value correction in multiple tests (57). Significant effects were interpreted only when of sufficient intensity (η²p > .02). All values are presented as mean (SD) unless stated otherwise. In the absence of three-level interaction effects, only two-level interaction effects were reported for each factor combination. Note that the degrees of freedom of the analysis are varied across variables due to differing exclusion rates for subjects.
Tasks performance and kinematics
The movement parameters for the circular steering task (IPe, speed, accuracy, range of trunk anterior flexion, range of elbow extension) were evaluated through a mixed ANOVA, which included group (healthy and stroke) as a between-subject factor, and hand (non-paretic/dominant and paretic/non-dominant hand) as a within-subject factor. Similarly, a mixed ANOVA was employed for the paced-reaching task (PANU, mean velocity, range of trunk anterior flexion, range of elbow extension), incorporating group (healthy and stroke) as a between-subject factor and hand (non-paretic/dominant and paretic/non-dominant hand) and condition (spontaneous- SAU and maximal- MAU) as within-subject factors.
Cortical activations
For the analysis of fNIRS peak of ΔHbO2 and fEEG Alpha and Beta ESRPs, a mixed ANOVA was applied with group (healthy and stroke) as a between-subject factor, and hand (non-paretic / dominant and paretic / non-dominant hand), condition (spontaneous- SAU and maximal- MAU, paced-reaching task), and hemisphere (contralateral / ipsilesional ; ipsilateral / contralesional) as within-subject factors.
Brain-movement relationship
In our investigation of the association between performance in the circular steering task and brain activation (fNIRS peak ΔHbO2) across the groups, we consistently applied Spearman rank correlation analysis. This approach was chosen to account for the non-normal distribution of some variables and to maintain consistency across the analysis, thus enhancing comparability of our findings. We choose to keep only moderate effects to avoid false effects, thus, we just present correlation with at least a p < .01 and a rs² > .25. Only those effects were reported to facilitate the results presentation.