MetaBCI: An open-source platform for brain-computer interfaces.

BACKGROUND: Recently, brain-computer interfaces (BCIs) have attracted worldwide attention for their great potential in clinical and real-life applications. To implement a complete BCI system, one must set up several links to translate the brain intent into computer commands. However, there is not an open-source software platform that can cover all links of the BCI chain. METHOD: This study developed a one-stop open-source BCI software, namely MetaBCI, to facilitate the construction of a BCI system. MetaBCI is written in Python, and has the functions of stimulus presentation (Brainstim), data loading and processing (Brainda), and online information flow (Brainflow). This paper introduces the detailed information of MetaBCI and presents four typical application cases. RESULTS: The results showed that MetaBCI was an extensible and feature-rich software platform for BCI research and application, which could effectively encode, decode, and feedback brain activities. CONCLUSIONS: MetaBCI can greatly lower the BCI's technical threshold for BCI beginners and can save time and cost to build up a practical BCI system. The source code is available at https://github.com/TBC-TJU/MetaBCI, expecting new contributions from the BCI community.

Forensic Study on Objective Evaluation of Visual Acuity of Ametropia with the Event-related Potential P3.

OBJECTIVE: In this study, we aimed to assess the characteristics of the P3 component from an event-related potential (ERP) that was induced by visual acuity (VA) processing. Furthermore, we sought to provide electrophysiological evidence for the objective evaluation of VA. METHODS: We recruited 32 participants with myopia-related ametropia. They reported no other ocular diseases and had an uncorrected VA of 4.0 in both eyes. We used the block letter "E" at different visual angles and orientations as the graphic stimuli. The oddball paradigm, consisting of 4 modules, was used for ERP analysis. The standard stimuli of each module were identical, with a visual angle of 1°15'. The visual angles of the target stimuli were 1°15', 55', 24', and 15'. The VA test was performed on each eye separately for all participants, and all characteristics of the P3 component were analyzed. RESULTS: There was no significant difference in the P3 peak letencies between the target stimulation angle 1°15' group and the 55' group, or between the target stimulation angle 24' group and the 15' group. There was a significant difference in the P3 peak letencies between the target stimulation angle 1°15' group and the 24' group as well as the 15' group. There was a significant difference in the P3 peak letencies between the target stimulation angle 55' group and the 24' group as well as the 15' group. No significant differences were observed in the P3 amplitude between modules. CONCLUSION: In the oddball paradigm, P3 elicitation indicated a cognitive response to the target stimuli. These data showed that the characteristics of P3 can be used as an objective evaluation of VA.

Single-trial P300 classification algorithm based on centralized multi-person data fusion CNN.

Introduction: Currently, it is still a challenge to detect single-trial P300 from electroencephalography (EEG) signals. In this paper, to address the typical problems faced by existing single-trial P300 classification, such as complex, time-consuming and low accuracy processes, a single-trial P300 classification algorithm based on multiplayer data fusion convolutional neural network (CNN) is proposed to construct a centralized collaborative brain-computer interfaces (cBCI) for fast and highly accurate classification of P300 EEG signals. Methods: In this paper, two multi-person data fusion methods (parallel data fusion and serial data fusion) are used in the data pre-processing stage to fuse multi-person EEG information stimulated by the same task instructions, and then the fused data is fed as input to the CNN for classification. In building the CNN network for single-trial P300 classification, the Conv layer was first used to extract the features of single-trial P300, and then the Maxpooling layer was used to connect the Flatten layer for secondary feature extraction and dimensionality reduction, thereby simplifying the computation. Finally batch normalisation is used to train small batches of data in order to better generalize the network and speed up single-trial P300 signal classification. Results: In this paper, the above new algorithms were tested on the Kaggle dataset and the Brain-Computer Interface (BCI) Competition III dataset, and by analyzing the P300 waveform features and EEG topography and the four standard evaluation metrics, namely Accuracy, Precision, Recall and F1-score,it was demonstrated that the single-trial P300 classification algorithm after two multi-person data fusion CNNs significantly outperformed other classification algorithms. Discussion: The results show that the single-trial P300 classification algorithm after two multi-person data fusion CNNs significantly outperformed the single-person model, and that the single-trial P300 classification algorithm with two multi-person data fusion CNNs involves smaller models, fewer training parameters, higher classification accuracy and improves the overall P300-cBCI classification rate and actual performance more effectively with a small amount of sample information compared to other algorithms.

Attention-based 3D convolutional recurrent neural network model for multimodal emotion recognition.

Introduction: Multimodal emotion recognition has become a hot topic in human-computer interaction and intelligent healthcare fields. However, combining information from different human different modalities for emotion computation is still challenging. Methods: In this paper, we propose a three-dimensional convolutional recurrent neural network model (referred to as 3FACRNN network) based on multimodal fusion and attention mechanism. The 3FACRNN network model consists of a visual network and an EEG network. The visual network is composed of a cascaded convolutional neural network-time convolutional network (CNN-TCN). In the EEG network, the 3D feature building module was added to integrate band information, spatial information and temporal information of the EEG signal, and the band attention and self-attention modules were added to the convolutional recurrent neural network (CRNN). The former explores the effect of different frequency bands on network recognition performance, while the latter is to obtain the intrinsic similarity of different EEG samples. Results: To investigate the effect of different frequency bands on the experiment, we obtained the average attention mask for all subjects in different frequency bands. The distribution of the attention masks across the different frequency bands suggests that signals more relevant to human emotions may be active in the high frequency bands γ (31-50 Hz). Finally, we try to use the multi-task loss function Lc to force the approximation of the intermediate feature vectors of the visual and EEG modalities, with the aim of using the knowledge of the visual modalities to improve the performance of the EEG network model. The mean recognition accuracy and standard deviation of the proposed method on the two multimodal sentiment datasets DEAP and MAHNOB-HCI (arousal, valence) were 96.75 ± 1.75, 96.86 ± 1.33; 97.55 ± 1.51, 98.37 ± 1.07, better than those of the state-of-the-art multimodal recognition approaches. Discussion: The experimental results show that starting from the multimodal information, the facial video frames and electroencephalogram (EEG) signals of the subjects are used as inputs to the emotion recognition network, which can enhance the stability of the emotion network and improve the recognition accuracy of the emotion network. In addition, in future work, we will try to utilize sparse matrix methods and deep convolutional networks to improve the performance of multimodal emotion networks.

Corrigendum: Multi-person feature fusion transfer learning-based convolutional neural network for SSVEP-based collaborative BCI.

[This corrects the article DOI: 10.3389/fnins.2022.971039.].

Evaluation of Visual Acuity of Ametropia with Visual Event-Related Potential Nogo-P3 Component.

OBJECTIVES: To analyze the Nogo-P3 component of event-related potential (ERP) in the process of visual acuity processing, to provide electrophysiological evidence for objective evaluation of visual acuity. METHODS: Twenty-six subjects with no other ocular diseases except for ametropia were recruited, and all subjects had uncorrected visual acuity both eyes 1/10 (evaluated using Monoyer chart). Block letter E with different visual angles and directions were used as graphic stimuli. The Go/Nogo paradigm was used for ERP studies. The visual angle of Go stimulation angle was 1°15', Nogo stimuli were 1°15', 55', 24' and 15'. The visual acuity test was performed on each of the two naked eyes separately in all subjects, and the characteristics of the Nogo-P3 component were analyzed. RESULTS: The latency of Nogo-P3 showed no difference between the stimuli of 1°15' and 55', and between Nogo stimulation angle 24' and 15'. There was significant difference between Nogo stimulation angle 1°15' and 24', and between Nogo stimulation angle 1°15' and 15' (P<0.05). There was significant difference between Nogo stimulation angle 55' and 24', and between Nogo stimulation angle 55' and 15' (P<0.05). No significant differences were observed in the Nogo-P3 amplitude among Nogo stimulation. CONCLUSIONS: In the Go/Nogo paradigm, Nogo-P3 can reflect the cognitive response of subjects to Nogo stimulation, which can be used for objective evaluation of visual acuity.

Multi-person feature fusion transfer learning-based convolutional neural network for SSVEP-based collaborative BCI.

Objective: The conventional single-person brain-computer interface (BCI) systems have some intrinsic deficiencies such as low signal-to-noise ratio, distinct individual differences, and volatile experimental effect. To solve these problems, a centralized steady-state visually evoked potential collaborative BCI system (SSVEP-cBCI), which characterizes multi-person electroencephalography (EEG) feature fusion was constructed in this paper. Furthermore, three different feature fusion methods compatible with this new system were developed and applied to EEG classification, and a comparative analysis of their classification accuracy was performed with transfer learning-based convolutional neural network (TL-CNN) approach. Approach: An EEG-based SSVEP-cBCI system was set up to merge different individuals' EEG features stimulated by the instructions for the same task, and three feature fusion methods were adopted, namely parallel connection, serial connection, and multi-person averaging. The fused features were then input into CNN for classification. Additionally, transfer learning (TL) was applied first to a Tsinghua University (THU) benchmark dataset, and then to a collected dataset, so as to meet the CNN training requirement with a much smaller size of collected dataset and increase the classification accuracy. Ten subjects were recruited for data collection, and both datasets were used to gauge the three fusion algorithms' performance. Main results: The results predicted by TL-CNN approach in single-person mode and in multi-person mode with the three feature fusion methods were compared. The experimental results show that each multi-person mode is superior to single-person mode. Within the 3 s time window, the classification accuracy of the single-person CNN is only 90.6%, while the same measure of the two-person parallel connection fusion method can reach 96.6%, achieving better classification effect. Significance: The results show that the three multi-person feature fusion methods and the deep learning classification algorithm based on TL-CNN can effectively improve the SSVEP-cBCI classification performance. The feature fusion method of multi -person parallel feature connection achieves better classification results. Different feature fusion methods can be selected in different application scenarios to further optimize cBCI.

Multitask network for thyroid nodule diagnosis based on TI-RADS.

PURPOSE: Assessment of thyroid nodules is usually relied on the experience of the radiologist and is time-consuming. Classification model of thyroid nodules cannot only reduce the burden on physicians but also provide objective recommendations. However, most classification models based on deep learning simply give a prediction result of the benignity or malignancy of nodules; thus, physicians have no way of knowing how the deep learning gets the prediction result due to the black-box nature of neural networks. In this work, we integrate the explainability directly into the outputs generated by the model through combining thyroid imaging reporting and data system (TI-RADS). The inference process of the proposed method is consistent with doctor's clinical diagnosis process; therefore, doctors can better explain the diagnosis results of the model to the patient. METHODS: A multitask network based on TI-RADS (MTN-TI-RADS) for the classification of thyroid nodules is proposed. In this network, a set of TI-RADS classifications of nodules is first obtained by multitask learning, then the TI-RADS points and the corresponding risk levels are calculated, and finally, nodules are classified as benign and malignant. The classification process through the network is consistent with the diagnostic process of physician; thus, the results of classification can be easily understood by physicians. In addition, the attention modules are introduced to the spatial and channel domains to let the network focus more on critical features. RESULTS: To verify the classification performance of our method, we compared the results obtained through our method with the results of the radiologist's evaluation. For the 781 test nodules in the internal dataset and the 886 test nodules in the external dataset, the sensitivity and specificity of MTN-TI-RADS were 0.988, 0.912 in internal dataset, 0.949, 0.930 in external dataset, versus the senior radiologist of 0.925 ( p < 0.001 $p<0.001$ ), 0.816 ( p = 0.005 $p=0.005$ ), and 0.910 ( p = 0.009 $p=0.009$ ), 0.836 ( p < 0.001 $p<0.001$ ), respectively. And the area under the receiver operating characteristic curve of MTN-TI-RADS was 0.981 in internal dataset, 0.973 in external dataset, versus the senior radiologist of 0.905, 0.923. For the internal dataset, we also computed the accuracy of the risk level (TR1 to TR5) and the mean absolute error (MAE). The accuracy of the risk level of the proposed method is 78%, and the MAE is 1.30. The MAE of the total points (0-14 points) is 1.30. CONCLUSIONS: An effective and result-interpretable end-to-end thyroid nodule classification network (MTN-TI-RADS) is proposed. MTN-TI-RADS has superior ability to classify malignant and benign thyroid nodules compared to senior radiologists. Based on MTN-TI-RADS, a classification model with strong interpretation and a high degree of physician trust is constructed. The proposed classification network is consistent with the diagnosis process of physicians, thus is more reliable and interpretable, and has great potential for clinical application.

Effect of Transcranial Direct Current Stimulation on the Mismatch Negativity Features of Deviated Stimuli in Children With Autism Spectrum Disorder.

Autism spectrum disorder (ASD) is a devastating mental disorder in children. Currently, there is no effective treatment for ASD. Transcranial direct current stimulation (tDCS), which is a non-invasive brain stimulation neuromodulation technology, is a promising method for the treatment of ASD. However, the manner in which tDCS changes the electrophysiological process in the brain is still unclear. In this study, we used tDCS to stimulate the dorsolateral prefrontal cortex area of children with ASD (one group received anode tDCS, and the other received sham tDCS) and investigated the changes in evoked EEG signals and behavioral abilities before and after anode and sham stimulations. In addition to tDCS, all patients received conventional rehabilitation treatment. Results show that although conventional treatment can effectively improve the behavioral ability of children with ASD, the use of anode tDCS with conventional rehabilitation can boost this improvement, thus leading to increased treatment efficacy. By analyzing the electroencephalography pre- and post-treatment, we noticed a decrease in the mismatch negativity (MMN) latency and an increase in the MMN amplitude in both groups, these features are considered similar to MMN features from healthy children. However, no statistical difference between the two groups was observed after 4 weeks of treatment. In addition, the MMN features correlate well with the aberrant behavior checklist (ABC) scale, particularly the amplitude of MMN, thus suggesting the feasibility of using MMN features to assess the behavioral ability of children with ASD.

Hydrogen sulfide restores cardioprotective effects of remote ischemic preconditioning in aged rats via HIF-1α/Nrf2 signaling pathway.

The present study explored the therapeutic potential of hydrogen sulfide (H2S) in restoring aging-induced loss of cardioprotective effect of remote ischemic preconditioning (RIPC) along with the involvement of signaling pathways. The left hind limb was subjected to four short cycles of ischemia and reperfusion (IR) in young and aged male rats to induce RIPC. The hearts were subjected to IR injury on the Langendorff apparatus after 24 h of RIPC. The measurement of lactate dehydrogenase, creatine kinase and cardiac troponin served to assess the myocardial injury. The levels of H2S, cystathionine ß-synthase (CBS), cystathionine γ-lyase (CSE), nuclear factor erythroid 2-related factor 2 (Nrf2), and hypoxia-inducible factor (HIF-1α) were also measured. There was a decrease in cardioprotection in RIPC-subjected old rats in comparison to young rats along with a reduction in the myocardial levels of H2S, CBS, CSE, HIF-1α, and nuclear: cytoplasmic Nrf2 ratio. Supplementation with sodium hydrogen sulfide (NaHS, an H2S donor) and l-cysteine (H2S precursor) restored the cardioprotective actions of RIPC in old hearts. It increased the levels of H2S, HIF-1α, and Nrf2 ratio without affecting CBS and CSE. YC-1 (HIF-1α antagonist) abolished the effects of NaHS and l-cysteine in RIPC-subjected old rats by decreasing the Nrf2 ratio and HIF-1α levels, without altering H2S.The late phase of cardioprotection of RIPC involves an increase in the activity of H2S biosynthetic enzymes, which increases the levels of H2S to upregulate HIF-1α and Nrf2. H2S has the potential to restore aging-induced loss of cardioprotective effects of RIPC by upregulating HIF-1α/Nrf2 signaling.

Comparison of ASSR, ABR and 40 Hz AERP Response Thresholds at Different Frequencies and Their Forensic Applications.

OBJECTIVES: To explore the relationship between the frequency characteristics and response threshold of auditory steady-state response (ASSR), auditory brainstem response (ABR) and 40 Hz auditory event related potential (40 Hz AERP), and their application values in forensic medicine. METHODS: Thirty volunteers with normal hearing (60 ears) were selected to perform pure tone audiometry (PTA) threshold and ASSR, ABR and 40 Hz AERP response threshold tests in the standard sound insulation shielding room, and the results were statistically analyzed by SPSS 22.0 software. RESULTS: At 0.5 kHz and 1.0 kHz frequencies, the correlation between 40 Hz AERP response threshold and PTA threshold was good, which was better than that of ASSR and ABR response threshold. At 2.0 kHz and 4.0 kHz frequencies, the correlation between ASSR and ABR response thresholds and PTA threshold was good, which was better than that of 40 Hz AERP response threshold. CONCLUSIONS: To evaluate the hearing at 0.5 kHz and 1.0 kHz frequencies, it is recommended to use 40 Hz AERP and ASSR to comprehensively assess the PTA threshold of the subjects. To evaluate the hearing at 2.0 kHz and 4.0 kHz frequencies, ABR and ASSR are recommended to assess the PTA threshold of subjects comprehensively. The combination of ASSR, ABR and 40 Hz AERP can improve the accuracy of hearing function evaluation.

Motor Imagination of Lower Limb Movements at Different Frequencies.

Motor imagination (MI) is the mental process of only imagining an action without an actual movement. Research on MI has made significant progress in feature information detection and machine learning decoding algorithms, but there are still problems, such as a low overall recognition rate and large differences in individual execution effects, which make the development of MI run into a bottleneck. Aiming at solving this bottleneck problem, the current study optimized the quality of the MI original signal by "enhancing the difficulty of imagination tasks," conducted the qualitative and quantitative analyses of EEG rhythm characteristics, and used quantitative indicators, such as ERD mean value and recognition rate. Research on the comparative analysis of the lower limb MI of different tasks, namely, high-frequency motor imagination (HFMI) and low-frequency motor imagination (LFMI), was conducted. The results validate the following: the average ERD of HFMI (-1.827) is less than that of LFMI (-1.3487) in the alpha band, so did (-3.4756 < -2.2891) in the beta band. In the alpha and beta characteristic frequency bands, the average ERD of HFMI is smaller than that of LFMI, and the ERD values of the two are significantly different (p=0.0074 < 0.01; r = 0.945). The ERD intensity STD values of HFMI are less than those of LFMI. which suggests that the ERD intensity individual difference among the subjects is smaller in the HFMI mode than in the LFMI mode. The average recognition rate of HFMI is higher than that of LFMI (87.84% > 76.46%), and the recognition rate of the two modes is significantly different (p=0.0034 < 0.01; r = 0.429). In summary, this research optimizes the quality of MI brain signal sources by enhancing the difficulty of imagination tasks, achieving the purpose of improving the overall recognition rate of the lower limb MI of the participants and reducing the differences of individual execution effects and signal quality among the subjects.

JAK/STAT and VEGF/PAK1 signaling as emerging targets for topical treatment of psoriasis: a pilot study.

Psoriasis is reportedly modulated by the Janus kinase (JAK)/signal transducer and activator of transcription (STAT) or vascular endothelial growth factor/p21-activated kinase 1 (VEGF/PAK1) pathways. However, no research has evaluated the expression of JAK/STAT and VEGF/PAK1 signaling pathway molecules in human psoriasis skin tissue concurrently. We investigated the expression of autocrine STAT1, STAT3, VEGF, suppressor of cytokine signaling-1 (SOCS1), SOCS3, and PAK1 in psoriatic tissues. Skin biopsies were retrospectively collected from 55 patients with psoriasis from the tissue biobank. Skin biopsies from 40 healthy volunteers undergoing plastic surgery were used as controls. Immunohistochemical staining revealed that STAT1, STAT3, SOCS1, SOCS3, VEGF, and PAK1 were present at significantly higher levels in the psoriasis samples compared to the control group. Similarly, the mRNA expression of these signaling molecules was also significantly upregulated in psoriatic skin. Additionally, some of the molecules in these two signaling pathways exhibited significant positive correlations. In summary, we present pilot evidence that JAK/STAT and VEGF/PAK1 signaling molecules are expressed in psoriasis, which may provide topical treatment targets for this disease.

Characteristics of Pattern Visual Evoked Potential in Two Eyes with Varying Visual Acuity in One Eye and Forensic Application.

In this study, we aimed to study the pattern visual evoked potentials (P-VEPs) in two eyes with varying visual acuity in one eye and to provide an objective estimation of visual acuity by comparing P-VEPs in one and two eyes. Thirty subjects were chosen, who had one eye with an acuity of 5.0, 4.85, 4.6, 4.0, or scieropia and obstructed vision and the other eye with an acuity of 5.0, respectively. P-VEPs were detected under the large grating stimuli at 3×4 spatial frequency, moderate grating stimuli (12×16 spatial frequency) and small grating stimuli (48×64 spatial frequency). Under large grating stimuli, there was no significant difference in P100 peak latency between the groups, nor was there a significant difference between the amplitude of two eyes and the amplitude of one normal-vision eye. Under moderate and small grating stimuli, there was a significant difference in P100 peak latency between the group with both eyes having an acuity of 5.0 and the group with visual acuity below 4.0 in one eye. There was a significant difference in P100 amplitude between the group with visual acuity of 5.0 in both eyes and the group with one normal-vision eye. There was no significant difference in the amplitude of two eyes and the amplitude of one normal-vision eye between any other two groups. In forensic identification, characteristics and variability of P-VEPs in one and two eyes can be used to identify malingering or decline in visual acuity.

Radial Basis Function Neural Network-based PID model for functional electrical stimulation system control.

Functional electrical stimulation (FES) has been widely used in the area of neural engineering. It utilizes electrical current to activate nerves innervating extremities affected by paralysis. An effective combination of a traditional PID controller and a neural network, being capable of nonlinear expression and adaptive learning property, supply a more reliable approach to construct FES controller that help the paraplegia complete the action they want. A FES system tuned by Radial Basis Function (RBF) Neural Network-based Proportional-Integral-Derivative (PID) model was designed to control the knee joint according to the desired trajectory through stimulation of lower limbs muscles in this paper. Experiment result shows that the FES system with RBF Neural Network-based PID model get a better performance when tracking the preset trajectory of knee angle comparing with the system adjusted by Ziegler- Nichols tuning PID model.

Electroencephalograph (EEG) signal processing method of motor imaginary potential for attention level classification.

Research of visual attention is one of the important domains of psychology and neurophysiology. In this study, an attention related electroencephalograph (EEG) signal processing method was proposed to distinguish the different levels of people's attention during the imaginary limbs motor. There were two EEG feedback experiments (playing tennis and walking) to measure the different levels of visual attention. Three imaginary motor tasks (attention, inattention, and rest task) were performed with the flash stimulus displayed on the screen in the experiments. A nonlinear dynamics parameter of multi-scale entropy (MSE) was extracted from those EEG data recorded. According to the statistics analysis of 14 subjects, there was an obvious declining tendency of MSE with the level of attention declining, which validated the effectiveness of the proposed method to classify the visual attention level.

Mu rhythm desynchronization detection based on empirical mode decomposition.

The aim of this paper is to investigate the possibility of using empirical mode decomposition (EMD) method in detecting the desynchronized mu rhythm of motor imagery EEG signal. A number of EEG studies have identified the mu rhythm desynchronization a reliable EEG pattern for brain-computer interface. Considering the non-stationary characteristics of the motor imagery EEG, the EMD method is proposed to decompose the EEG signal into intrinsic mode functions (IMFs). By analyzing the power spectral density (PSD) of the IMFs, the characteristics one representing mu rhythm oscillations can be detected. Then by Hilbert transformation, the event-related desynchronization phenomenon can be found by the envelope of the characteristics IMF. Results demonstrate that the EMD method is an effective time-frequency analysis tool for non-stationary EEG signal.

A gait stability investigation into FES-assisted paraplegic walking based on the walker tipping index.

The gait outcome measures used in clinical trials of paraplegic locomotor training determine the effectiveness of improved walking function assisted by the functional electrical stimulation (FES) system. Focused on kinematic, kinetic or physiological changes of paraplegic patients, traditional methods cannot quantify the walking stability or identify the unstable factors of gait in real time. Up until now, the published studies on dynamic gait stability for the effective use of FES have been limited. In this paper, the walker tipping index (WTI) was used to analyze and process gait stability in FES-assisted paraplegic walking. The main instrument was a specialized walker dynamometer system based on a multi-channel strain-gauge bridge network fixed on the frame of the walker. This system collected force information for the handle reaction vector between the patient's upper extremities and the walker during the walking process; the information was then converted into walker tipping index data, which is an evaluation indicator of the patient's walking stability. To demonstrate the potential usefulness of WTI in gait analysis, a preliminary clinical trial was conducted with seven paraplegic patients who were undergoing FES-assisted walking training and seven normal control subjects. The gait stability levels were quantified for these patients under different stimulation patterns and controls under normal walking with knee-immobilization through WTI analysis. The results showed that the walking stability in the FES-assisted paraplegic group was worse than that in the control subject group, with the primary concern being in the anterior-posterior plane. This new technique is practical for distinguishing useful gait information from the viewpoint of stability, and may be further applied in FES-assisted paraplegic walking rehabilitation.