그림 1. 상완 이두근 수축 및 이완 동안 근전도 (EMG) 원시 데이터.

Figure 2. (a) Image of surface EMG: the red and black lines represent the + and − electrodes, and the blue line represents the ground; (b) image of the needle EMG.

1.1. Surface EMG

Surface EMG is the most widely known and used method in measuring muscle contraction. Electrical contributions made by the active motor units are measured in a noninvasive way on the skin using electrodes [7]. sEMG does not simply measure the electrical signals generated by the muscles; it represents the potential difference between two electrodes. If two electrodes are placed along muscle fibers at both ends of the motor for the muscle, the same muscle action potentials at both ends of the muscle are recorded at the same time, making it impossible to measure the potential difference. Therefore, the EMG signal is measured using a differential amplifier that rejects a common signal from both electrodes using two electrodes [12]. Since electrical signals exist even in muscles that are not in use, reference electrodes are required to exclude redundant signals. Finally, three electrodes must be attached to receive the EMG signal of one muscle. Through this process, sEMG can evaluate the contractile ability of one muscle from three signals.

In the early days of sEMG, it was used for biofeedback. For example, a study was conducted to reduce the frequency and severity of tension headaches through EMG biofeedback, and another study found that the resting level of frontalis EMG activity is higher in tension headache patients than in the general population. From this, it was argued that contraction headaches caused by constant contractions of the scalp and neck muscles could be relieved by learning to relax these muscles through biofeedback [13]. In another study, when biofeedback was given through computerized electromyographic evaluation of pelvic floor muscles, the subjective evaluation of pain decreased by 83% after 16 weeks and the resting tension level decreased by 68% [14]. However, in the above study, since only the amplitude of the simple EMG value was used to check the contraction degree of the muscle and proceeded with biofeedback, more detailed EMG parameters were studied to evaluate the muscle performance.

To obtain the parameters, EMG data are mainly analyzed in two domains: a time domain where EMG data can be viewed over time and a frequency domain that can check the frequency of EMG data within a certain range. In the time domain, the degree of muscle activation can be visually confirmed over time. Because muscles require a larger electrical signal to generate more force, they exhibit a larger amplitude in the EMG data. Therefore, it is possible to check the degree of activation for each muscle using the root mean square (RMS) value [15,16]. Fate analysis in the time domain is also possible. When a muscle continuously exerts the same force, a larger electrical signal is required to maintain the same output; therefore, the more fatigue it builds up, the greater the amplitude displayed. Fatigue can be analyzed with an increase in RMS indicating an increase in amplitude [17]. In addition, various parameters such as integrated EMG and maximum amplitude are used in the time domain [18]. In the frequency domain, fatigue is mainly analyzed. As muscle fatigue builds up, the speed of transmitting electrical signals slows down and the frequency is lowered, so that the conduction velocity slows down [19]. To capture this, fatigue is quantified by using parameters such as mean frequency, median frequency, and peak frequency in a specific range [18]. An example of EMG parameters for fatigue analysis can be seen in Figure 3.

Figure 3. Surface EMG data obtained during voluntary contraction of a bicep with a 1-kg weight for a female: (a) raw data, (b) median frequency and linear regression, and (c) root mean square and linear regression.

Using these parameters, clinical studies have been conducted to measure muscle fatigue. Since the amplitude of EMG is different for each individual, fatigue was given using maximum voluntary contraction (MVC) and reliability was confirmed by analyzing the EMG parameters for fatigue (Table 1). Fatigue parameters were studied as RMS in the time domain, mean power frequency (MNF), and median power frequency (MDF) in the frequency domain, since the power spectrum frequency shift decreased. Mainly, the vastus lateralis (VL), rectus femoris (RF), vastus medialis (VM), lumbar 1–5, bicep brachii, and quadriceps were studied. Since the EMG signal has an accurate value when measured individually for one muscle rather than a complex signal of several muscles, it is limited to isometric contraction rather than dynamic contraction, which has a risk of complex measurement of several muscles. As a result of measuring EMG data during fatigue with isometric contraction, it was reported that most were reliable as indicators of fatigue, but due to the difficulty of application in dynamic contraction fatigue experiments and noise from electrode attachment, research on EMG continues.

sEMG has a limitation in that it is a noninvasive method that is not directly inserted into the muscle. Because of the noise caused by the electrode, which is an attachment type, and due to the muscle attachment position of the electrode, it affects the conduction velocity and median frequency, parameters of EMG [27]. Standardization of the attachment method and attachment location per laboratory was required. Therefore, a study on standardizing the electrode position was conducted to obtain accurate and repeatable data on the sEMG signal parameters. Recommendations for the use and technical considerations of sEMG and a questionnaire and interpretations required for the use of sEMG were presented [6]. Several researchers studied the standardized electrode locations. For example, to provide information on the degree of uniformity of the inner zone position of 13 superficial muscles in the lower limb, an experiment was conducted and the optimal electrode placement was between the inner zone and the tendon termination according to the landmark in eight of 13 muscles [28]. This placement was suggested by examining the muscle fiber orientation and palpable bony landmarks in the abdominal muscle [29]. The surface EMG for a Noninvasive Assessment of Muscles (SENIAM) project standardized the electrode attachment location and electrode size for concerted action in the Biomedical Health and Research program of the European Union [30]. However, even if the standard electrode location is used, the biggest drawback of sEMG is not compensated. The limitations from previous studies are presented in Table 2. Some obtained the same results as the isometric contraction in dynamic contraction, and some did not obtain valid data. The studies that did not obtain valid data from the isometric contraction suggested that the reason was due to the movement of various muscles rather than one muscle when performing a specific movement.

When the EMG sensor is applied to a smart device, it is possible to recognize and prevent a muscle from being injured during exercise by monitoring the movement of the muscle. Since 70 years have passed since it was used for research, if a wearable sensor is made using the results of the research conducted so far, quality of life will be greatly improved. For example, there was a study that analyzed the frequency spectrum of electrocardiography (ECG) data that can detect EMG and EEG and can differentiate sleep states divided into four with 98–99% accuracy [37]. In addition, some studies have shown that driver fatigue can be measured by ECG and EMG [38,39]. In the case of drowsy driving, ECG and EMG can be developed for wearable biosensors to detect the risk of drowsiness at one of four levels and can greatly reduce accident rates. However, since sEMG and EEG need to have electrodes attached, they are inconvenient to wear and have not yet been commercialized due to noise problems. It was mainly analyzed only in the isometric contraction of a single muscle, and since dynamic movement appears when the movements of several muscles are integrated, it is difficult to measure accurate and significant data when performing dynamic movements. Also, a wireless sensor is essential for use outside the laboratory, such as running on a track or in a playground. Currently, a wireless sEMG sensor has been developed to analyze dynamic motion, but the noise that occurs in dynamic motion has not yet been resolved. For the EMG sensor to become a wearable sensor, it is necessary to solve the problem of noise generated on the surface electrode. The solution to this will be covered in detail in the section on noise below.

1.2. Needle EMG

Unlike surface EMG, needle EMG records the electrical activity of a muscle by inserting a needle electrode directly into the muscle or by obtaining the high-frequency electrical activity generated by moving the position of the needle electrode in the muscle and an electrical signal in a rest state [40]. The measurement of needle EMG by physical stimulation causes more severe pain than that of sEMG. In order to solve this problem, a study was conducted to investigate the pain of needle EMG [41], which can be divided into two types: a concentric needle electrode and a monopolar needle electrode (Figure 4). In the former, the surrounding cannula is the reference electrode, and the latter is recorded through the surface electrode. The electrode for measuring needle EMG is mainly a concentric needle electrode that has a small recording area and can obtain a value that cancels out noise. Since needle EMG can measure accurate data, one unit of motion can be analyzed through action potentials obtained from muscle fibers that contract at a location very close to the needle electrode using triggered averaging and decomposition methods [42]. After inserting a needle to induce minimal muscle contraction, the motion unit potential is analyzed by continuously obtaining one and the same motion unit potential using amplitude triggering while gradually moving the position. Muscles are evaluated mainly by comparing the average duration and amplitude with normal values [43]. After that, multi-motor unit potentials (MUP) analysis was developed and a technique capable of simultaneously obtaining and analyzing multiple-MUP from one recording site became available and was applied [44]. The pain of needle EMG, which was caused by the movement of the existing electrode, is less than in the past with multi-MUP analysis, which is possible with one insertion. However, there are still various limitations.

Figure 4. (a) Monopolar needle electrode; (b) concentric needle electrode.

Electrodiagnostic physicians should conduct a needle EMG test after checking the patient’s history based on clinical data. Care should be taken regarding some patients with skin infections, skin diseases, bleeding disorders, and obesity. Since there is a risk due to various variables such as muscle location and muscle size, needle EMG requires that the placement of a needle in the longitudinal midline of the muscle is accurately inserted into the muscle of interest. Instead of these drawbacks, it can be used as a measure of the accuracy of sEMG. For example, in the vastus intermedius muscle, sEMG was used to determine if the whole muscle could be used to assess neuromuscular activation. The observation of a good correlation could be used to assess total neuromuscular activation of the VL muscle during isometric contraction at low force levels [45]. There is also a paper on the feature extraction of a forearm EMG signal for prosthetics. EMG data analysis was conducted using both needle and sEMG for recording accuracy during hand movement [46]. For analysis of neuromuscular jitter that can be measured only in a single fiber, only needle EMG can be used [47,48,49] (Figure 5). EMG, which can obtain sensitive signals, is used for physiological analysis and diagnosis in a single motor unit during muscle contraction. Needle EMG is known to have a more accurate value than sEMG and is being used and studied. Needles have little noise, but considering the risk and difficulty of including them in a wearable sensor, it cannot be applied to the future development of biosensors. However, if the surface electrode is made in a way that reduces noise by bringing the reference electrode closer like a needle EMG to make it more accurate and easy to wear, it can be used in a wider range.

그림 5. 혀의 바늘 EMG : 건강한 피험자의 오른쪽 genioglossus 근육에서 모터 단위 활동 전위 (MUAP). MUAP 기간에 대한 마커는 5 개의 평균 MUAP 중 가장 오염되지 않은 것으로 설정되었습니다. BMJ Publishing Group Ltd.의 허가를 받아 [ 50 ] 에서 복제 됨 .