| The restoration of motor functions of paralyzed patients is a critical issue of rehabilitation medicine. The current methods of motor function rebuilding are mainly based on either biomedicine or engineering technology. The neuromuscular electrical stimulation (NMES) is a commonly used electrical engineering method for paralysis rehabilitation. NMES elicits the contraction of target muscles by applying stimulation current to the targeted peripheral neuromuscular system. It is well adopted because of the convenience and effectiveness. However, the current NMES systems are usually lack of controlling flexibility and voluntary participation of patients. In addition, due to the inflexible stimulation parameters, the existing NMES methods are likely to cause a rapid onset of muscle fatigue.In this study, we developed a strategy based on the communication principle, the microelectronic technology, and the electromyographic (EMG) control. The surface EMG (sEMG) signals of the agonist muscles were transformed to stimulation pulses with various parameters to stimulate the target paralyzed muscles and motion pattern recognition was included to recognize the motion status and mapping the sEMG detection channels to the corresponding stimulation channels. In contrast to the existing EMG-controlled NMES systems which generate previously determined stimulation patterns according to the EMG patterns, our method established an information bridge between the muscles of a healthy limb and a paralyzed limb. So it is called "Electromyographic Bridge" (EMGB). The test results showed that the EMGB system can regenerate the target movements with a higher fidelity and a lower fatigability, thus the voluntary control of patient can be significantly enhanced.The specific works of this paper is presented as follows:1) Because the amplitude of the informative propagating components of the sEMG is positively related to the distance between the source muscle fibers and the electrodes, while the non-propagating components of the sEMG have non-decayed property and the non-propagating componentsmainly contributes to the cross-talk interference, we plot a sEMG root mean square (RMS) map by using single differential electrodes array to facilitate the sEMG detection site selection of different agonist muscles. using2) The optimized stimulation sites of four movements (grasping, finger extension, wrist flexion and wrist extension) were studied, because they are most commonly used in clinical upper-extremity rehabilitation. In addition, the coordinates of the stimulation sites were described by using the proportional bone measurement of traditional Chinese medicine. Because the coordinates can be used across individuals, it is convenient for clinical practice.3) A detection to stimulation channel mapping algorithm based on motion pattern recognition was proposed. The class separability, classification accuracy and calculation time using four features (time domain, autoregressive coefficient, sample entropy and integrated bispectral feature) and two classifier (linear discriminant analysis and artificial neural network) between our RMS detection site selection method and traditional method were compared. Finally, it was found that by using the RMS site selection method, time domain features and linear discriminant analysis, appropriate balance between classification accuracy and calculation complexity can be achieved.4) The stimulation generation algorithm was studied. Firstly, the primary parameter "refractory period" of the stimulation generation algorithm formerly proposed by our group was analyzed using continuous wavelet singularity analysis. The linear relationship between force generation and stimulation frequency can be achieved by the determined refractory period. Then, considering the limitation of unchanged stimulation intensity in the previous algorithm, we proposed a stimulation pulse width and frequency co-modulation algorithm based on sEMG time domain features (mean absolute value, MAV and number of slope sign changed, NSS) called MNDC algorithm. The test results showed that the MNDC algorithm is more effective for both voluntary muscle force reproduction and muscle fatigue reduction than traditional pulse-width modulation strategy.5) For circuit design, the sEMG detection circuit was firstly improved according to the sEMG frequency characteristics. A multichannel, arbitrary-output, isolated current stimulator was also developed to output the stimulation signals. Based on the core circuits and algorithms mentioned above, four types of prototype EMGB system were developed including "two channel EMGB system", "EMGB system for finger motor function rebuilding", "multichannel EMGB system based on Lab VIEW and DSP", and "embedded multichannel EMGB system". The system performance were also tested on the healthy people and the test results showed the movement of the controller’s hand can be regenerated on the controlled hand with high accuracy and low latency.6) To verify whether EMGB system is benefit for the upper-extremity rehabilitation of hemiplegic patients after stroke by enhancing the voluntary participation of NMES training, a randomized controlled trial was conducted in Zhongda Hospital. Thirty six acute or sub-acute stroke patients were randomized to EMGB experimental group (n=18) and cyclic NMES control group (n=18). For each group, twenty days of EMGB or cyclic NMES training in addition to the ordinary clinical rehabilitation program were conducted. The evaluation results showed that, after the same amount of training, the upper-extremity Fugl-Meyer score, motor status score, sEMG-ratio, and active range of motion were significantly better for EMGB group than the cyclic NMES control group. Thus the rehabilitation effectiveness of EMGB method was proved for early-stage hemiplegic patients. |
PDF Full Text Request