Bilateral upper limb robot-assisted rehabilitation improves upper limb motor function in stroke patients: a study based on quantitative EEG

This study was conducted at the Affiliated Zhongda Hospital of Southeast University from 01 April 2021 to 28 February 2022. 24 patients were randomly assigned to two groups (Conventional group and Bilateral group) in a 1:1 ratio using a simple randomization procedure. The protocol was approved by the Committee of Institutional Ethics (No:2021ZDSYLL091-P01) and registered in the China Clinical Trial Registration Center (ChiCTR2100049484). All patients signed the informed consent.

The inclusion criteria were as follows: (i) diagnosis of stroke was confirmed by computed tomography (CT) and/or magnetic resonance imaging (MRI); (ii) aged 18–80 years old, clear consciousness, able to complete the corresponding assessment (Mini-Mental State Examination (MMSE) ≥ 20 points); (iii) first onset, course of 2 weeks to 3 months; (iv) the modified Ashworth spasticity assessment of the upper limb major muscle tension was less than or equal to grade 2; (v) Brunnstrom assessment of hemiplegic upper limb was in stage II–IV; (vi) stable clinical condition and good control of chronic diseases; (vii) the treatment methods of this study were understood and cooperated. The criteria for exclusion were (i) previous history of brain injury, brain tumor or other neurological diseases; (ii) pain or dysfunction of both upper limbs caused by other reasons; (iii) patients with severe diseases including the cardiovascular system, cerebrovascular system, digestive system, endocrine system, and psychosis; (iv) lesions in cerebellum or brainstem.

The bilateral upper limb robot-assisted system consists of two upper limb robot-assisted devices (Burt, ESTUN Inc., Nanjing, China) connected by a connecting cable and a computer display providing feedback. The devices are end-driven and can provide movement in three dimensions. They offer various modes of active, passive, weight loss assisted, and direction-restricted active–passive movement, with movement feedback for each mode through a real-time situational interactive game.

All patients of two groups received a 60 min routine rehabilitation treatment, including physical therapy (neuromuscular electrical stimulation, rolling, transferring, sitting, standing and walking) and occupational therapy (dressing, eating, grooming). Based on the routine treatment, the BRT group received 30 min of bilateral robot-assisted training, while the CT group received 30 min of task-related training, including reaching to move a cup, grasping and releasing blocks, picking up coins, and rolling a bucket. All treatments were given once a day, 6 days /week, for a total of 3 weeks.

For the bilateral upper limb robot-assisted therapy, the patient is seated in the treatment chair between two upper limb rehabilitation robots. The robots are then adjusted so that the blue line on the robot arm is aligned with the patient’s acromion. The upper limbs are secured to the end-effectors of the intelligent force feedback rehabilitation robot with glove straps to prevent slippage during treatment, and to secure the trunk. The training program was switched to bilateral arm training mode. The assistance force was adjusted to resist the gravity of arms. Then, the range of motion is calibrated according to the patient's active mobility or the therapist's prescription, and games with simultaneous bilateral movements are selected (drumming, flag waving, rowing), 30 min, respectively (Fig. 1).

All outcome measures were assessed pre-(T1) and post-treatment (T2; at the end of the 3 week treatment). The primary outcome measure was the Fugl-Meyer assessment-Upper Extremities (FMA-UE). Secondary outcomes included the modified Barthel index (MBI) score, as well as quantitative electroencephalography (QEEG). The observation indexes of QEEG include power of electroencephalogram rhythms and brain functional connectivity.

FMA-UE was used to evaluate upper limb motor function. There were 33 assessment sub-items with a total score of 66 points. The score standard as follows: 0 indicates completely inactive; 1 indicates only part of the activities can be completed; 2 indicates the ability to perform activities normally. The higher the score, the better motor function.

The MBI score was used to evaluate the ability to perform activities of daily living (ADL) including 10 functions, controlling urination, modification, toilet use, eating, bed and chair transfer, bathing, walking activities, dressing, and going up and down stairs. Full score = 100 points. A higher score reflects better independence, with a lower score reflecting stronger dependence.

The 64-lead EEG system of the EGI company in the United States was used for EEG acquisition. A quiet, semi-dark, and electrically shielded examination room was selected. Subjects were asked to keep awake, quiet, closed eyed, and comfortably sitting after breakfast in the morning with tests completed by the same EEG examiner. In the recording of EEG, the Cz electrode in the center of the head was used as the reference electrode, and the sampling frequency was 1000 Hz. The impedance of each channel was lower than 5 kΩ, and the 50 Hz power frequency interference was filtered for five minutes. The data processing and analysis software Matlab2013a of THE MATH WORKS company and EEGLAB toolkit were used to analyze EEG data offline. The reference electrode was converted from the recorded Cz electrode to the average of all electrodes, and the band-pass filter was 0.05–30 Hz; segmentation was carried out in 3 s. An independent component analysis (ICA) algorithm was used to correct artifacts such as electrocardiogram and EMG. Hamming window and fast Fourier transform (FFT) were used to perform spectrum analysis on non-overlapping windows of EEG records obtained during the resting state. In the frequency spectrum density spectrum, power amplitude is represented by the square of each Hertz microvolt, and the absolute power is analyzed by FFT transform to obtain the frequency characteristics of different frequency bands. EEG frequency band division is as follows: δ (1.0–3.9 Hz), θ (4.0–7.9 Hz), α (8.0–13.9 Hz), β (14.0–30.0 Hz). Functional connectivity (FC) uses the HERMES toolkit to calculate a weighted phase-lag index (wPLI) of α and β. The calculation method of wPLI is as follows: Frequency bands of interest (alpha, beta) are defined. The EEG data of electrode X and electrode Y were filtered according to the above interested frequency bands. The filtered signals of electrode X and Y were transformed with a Hilbert transform, and the phase φx and φy at each time point were extracted. Finally, the cross spectrum was calculated, and the wPLI value of each segment was calculated for the average between segments. Among them, the wPLI value range was 0–1, the larger the value, the stronger the functional connection (Fig. 2).

SPSS23.0 statistical software was used for statistical analysis. The measurement data with normal distribution and homogeneity of variance were expressed as (‾x ± s). The paired sample t-test was used for intra-group comparison, while the independent sample t-test was used for inter-group comparison. For count data that do not conform to normal distribution and homogeneity of variance, a Chi-square test was used, and the significance level was p < 0.05.

There were no significant differences in age, gender, course of disease, stroke type, or affected side between the two groups (Table 1).

As presented in Fig. 3, no significant differences in FMA-UE and MBI scores among the two groups before treatment existed (p > 0.05). Within-group differences in FMA-UE and MBI scores for both the bilateral and conventional groups were statistically significant after treatment. A between-group comparison indicated that the MBI score of the bilateral group was significantly higher than that of the conventional group, while the FMA-UE score was not significantly different from that of the conventional group. The differences of FMA-UE and MBI scores before and after treatment in the bilateral group was significantly different compare to those in the conventional group.

The differences in the density of current sources at different frequencies between the two groups after treatment were observed, and the topographic maps of the statistical results are shown in Figs. 4, 5. From the alpha frequency map (Fig. 4), it can be seen that localized brain activity in the CT group after treatment was in the frontal lobes (BA4, BA6, BA9) (blue part in Fig. 4A); the sites of enhanced activity in the bilateral group after treatment were located in the paracentral lobule (BA2, BA3) and precentral gyrus (BA4, BA6) (blue part in Fig. 4B). The sites of enhanced activity in the BRT group compared with the CT group were located in the precentral gyrus (BA4, BA6) (blue part in Fig. 4C). As seen from the β-frequency diagram (Fig. 5), the localized brain activity in the CT group was in the parietal lobe (BA39) after treatment (blue part in Fig. 5A), while the enhanced activity in the BRT group was in the frontal lobe (BA4, BA6) and parietal lobe (BA7, BA40) bilaterally (blue part in Fig. 5B). The enhanced activity in the bilateral group was in the frontal lobe (BA4) and parietal lobe (BA5) bilaterally compared with the CT group (blue part in Fig. 5C).

By comparing the brain functional connectivity constructed for the cortical regions of interest, it was found that at α frequency, the regions of interest (ROI) were BA4, BA6, BA7, BA40, BA31, and BA24. The brain functional connectivity was significantly different at the local frontal, limbic, and frontal–parietal connections (Fig. 6C), and this significant functional connectivity was more pronounced in the BRT group compared to the CT group. At β frequency, the regions of interest (ROI) were BA3, BA5, BA6, BA7, BA8, BA9, BA13, BA30, BA31, and BA40. There were significant differences in local parietal, frontal, and brain functional connectivity at local frontal, limbic, and frontal–limbic connections (Fig. 6F). This significant functional connectivity was also reflected in the cortical response after treatment in the BRT group.