Contra-lateral desynchronized alpha oscillations linearly correlate with discrimination performance of tactile acuity

Four unilateral forearm amputees (denoted as A1 to A4) and thirteen able-bodied subjects (10 males and 3 females, mean age ± SD = 24 ± 2, denoted as H1 to H13) participated in this study. All the amputees were right-handed before the amputation. None of the amputees had a history of neurological or psychiatric disease. All the subjects were paid for their participation and full written informed consents were obtained from them in accordance with the Declaration of Helsinki.

Before the experiment, finger-projected areas were identified on the stump for each amputee through palpation (touch the forearm stump using a test stylus with a globular head of 2 mm in diameter) [28]. Detailed characteristics for each amputee were shown in table 1.

The objective of this experiment was to investigate the brain response during tactile discrimination task and characterize the correlation between task-related EEG features and tactile discrimination performance, which could be used to elucidate the feasibility of EEG features as an objective index for tactile acuity evaluation. Subjects were required to complete two sessions (i.e. the vibration session and the pressure session) of tactile discrimination task. As a stimulus, multi-pulse train of biphasic, rectangular current pulses were generated by the Master-9 Pulse Stimulator with isolators (Iso-Flex, A.M.P.I Company, Israel). Sensations of vibration and pressure were generated by grading the frequency and pulse width, respectively. A kind of custom-made surface electrode array (Shanghai Global Biotech Inc. China) with two copper circles (5 mm in diameter) was used in this experiment. Stimulation was applied to the thumb and index for healthy hands, while the thumb and index-projected areas were stimulated for amputated hands. To reduce artifacts induced by electrical stimulation, each subject wore a wrist strap. The wrist strap was made of copper and connected to the ground of the stimulation electrical circuit [29]. Within the tactile discrimination task, subjects had to discriminate three stimulated areas and three levels of intensities (low/medium/high) in each session. In the vibration session, the pulse width was 200 µs. And frequencies were 20, 30 and 40 Hz for three levels of vibration intensities, respectively. In the pressure session, the frequency was set to 30 Hz. And pulse widths were 150 µs, 200 µs and 250 µs for three levels of pressure intensities, respectively. Detailed stimulation parameters were displayed in table 2 and table 3. The selection of frequencies in the vibration and pressure sessions were based on the results of our pilot test [27]. And all stimulation parameters in this experiment were all below the pain threshold which have been quantified in our previous work [28]. The stimulated side, the three stimulation levels as well as the three stimulated areas, were pseudorandomly intermixed for each session. Each session consisted of 14 types, and each type included 15 trials, yielding a total of 210 trials, as illustrated in figure 2(A). The order of session presentation was counterbalanced across subjects. Short breaks were provided to each subject between two sessions to avoid mental fatigue and habituation.

Zoom In Zoom Out Reset image size

EEG signals were simultaneously recorded during the tactile discrimination task. Subjects were seated in a comfortable armchair with both forearms and fingers resting on the armrest. In order to reduce movement-related artifacts, subjects were required to limit their eye blinks, arm and finger movements during stimulation. A liquid crystal display (LCD) was set approximately 60 cm in front of subjects to display visual cues. Prior to EEG recording, all subjects were familiarized with the tactile discrimination task and received about 50 trials with immediate feedback until they reached a stable performance level. The experimental protocol was illustrated in figure 2(B). At the beginning of each trial, a white cross with black background appeared on the LCD to remind the subject to be prepared for the subsequent tasks. At the 4th second, the electrotactile stimulation began and lasted for 4 s. Subjects were required to carefully feel the stimulation and try to judge the current stimulation type. Once they confirmed the specific stimulation type, they were required to step on the pedal and then just pay attention to feel the stimulation rather than re-discriminate the stimulation. The time ranged from stimulation onset to pedal stepping was regarded as stimulation judgement period in this study. When the stimulation stopped, they were asked to report the stimulation type (stimulated side: left or right; stimulated area: thumb, index or both; stimulation level: high, middle or low) that they had confirmed at the point of pedal stepping. Finally, a random time of 3–5 s was appended to the relaxation period to limit subjects' adaption. After the experiment, all amputees reported no stimulus-evoked pain on the stump or adjacent areas of the limb during the stimulation.

EEG signals were recorded using a SynAmps2 system (NeuroScan, U.S.A.) and a quick-cap with 64 Ag/AgCl electrodes. Electrodes were placed according to the extended 10-20 system. The reference electrode was located on vertex, and the ground electrode was located on forehead. The low-pass filter setting was 0.5–100 Hz with a sampling rate of 1000 Hz, and a 50 Hz notch filter was used to diminish power line interference. Impedances of all electrodes were kept below 5 kΩ.

The EEG data were segmented into 6 s epochs including [–3.5−0.5] s prior to stimulation onset (baseline) and [0.5 3.5] s post stimulation onset (task state). Trials contaminated by artifacts such as swallowing and movement were excluded (the ratio of rejected trials was 6 %). We defined EEG signals in the baseline and task state as two class labels, respectively. The R² index (squared Pearson-correlation coefficient between EEG features and class labels) [30] was used to evaluate the contribution of different frequency bands and EEG channels to the task-related response. It was calculated based on the spatial-spectral-temporal structures of EEG signals in the baseline and task state. With topographic R² value distributions, we can have better understanding for the related channel and frequency band of the brain activation during the tactile discrimination task. Here, the topoplot of R² values were averaged in the frequency band of alpha band (8–13 Hz).

As a variant of ERD, event-related spectral perturbation (ERSP) was used to visualize the spectral power changes corresponding to tactile discrimination task [31–33]. ERSP was generally formulated as

$$\begin{equation} ERSP(f,t) = 1/n\sum_{k = 1}^{n}(F_{k}(f,t)^{2}) \end{equation} \tag{ 1 }$$

where n was the number of trials, and F_k(f, t) was the spectral estimation of the k^th trial with frequency = f and time = t. Short-time Fourier transform (STFT) was applied in time-frequency analysis of EEG data. ERSP was calculated every 200 ms with a Hanning-tapered window, convoluted with a modified sinusoid basis in which the number of cycles linearly changed with frequencies to achieve proper time and frequency resolution [34]. Note that the ERSP values were log-transformed prior to further analysis.

To visualize the topographical distributions of cortical activations during the task, we divided the task period into two sequential time periods. The first period was chosen from stimulation onset to the pedal stepping, thus including the stimulation presentation and judgement, which was termed as stimulation judgement period. The second period was defined from the pedal stepping to the end of stimulation, in which only stimulation presentation was included. Trials in which the pedal was stepped after the end of stimulation were excluded in this analysis (the ratio of excluded trials was 7 %). In order to obtain an overall topographical distribution of cortical activation for the two sequential time periods, the spectral powers were averaged in the alpha band across subjects and all stimulation types of trials.

Based on topographic R² value distributions, channels with R² > 0.03 were chosen for the calculation of alpha ERD/ERS. There were two kinds of alpha ERD/ERS in this study, i.e. prefrontal alpha ERD/ERS and somatosensory alpha ERD/ERS. Here, channels of FP1, FPZ, FP2, AF3 and AF4 over the prefrontal cortex were selected to calculate the prefrontal alpha ERD/ERS. Channels of FC3, C5, C3, C1 and CP3 were selected to calculate left-hemisphere somatosensory alpha ERD evoked by right side stimulation. Similarly, channels of FC4, C2, C4, C6 and CP4 were selected to calculate right-hemisphere somatosensory alpha ERD/ERS evoked by left side stimulation. In the following description, we collectively called them contralateral somatosensory alpha ERD/ERS. ERSP values were normalized by subtracting the baseline of t = [–3.5−0.5] s for each of those channels. Alpha ERD/ERS was obtained by averaging the normalized ERSP values in alpha band (8–13 Hz) over corresponding period and channels.

For additional analysis of performance-related difference, an equal (i.e. the minimum) number of correct and incorrect trials were randomly selected for each subject and each stimulation type. The correct discrimination referred to subjects must report all correct information of the three kinds of evaluation types: intensity, stimulated area and stimulated side. Any trial with one or both two kinds information wrongly reported was considered as incorrect discrimination. The above analyses (topographical distributions of cortical activations and alpha ERD/ERS calculation) were then applied to the equal-numbered subsets of correct and incorrect trials, respectively.

In the stimulation discrimination task, accuracy rate (AR) and response time (RT) served as behavioral measures of tactile acuity. RT was the time from the beginning of the electrotactile stimulation to the onset of pedal stepping, which reflected the duration to discriminate the current stimulation type. AR was generally formulated as

$$\begin{equation} AR = \frac{Number \;of \;correctly \;identified \;trials}{Number \;of \;total \;trials}\times100\% \end{equation} \tag{ 2 }$$

In this study, AR/RT for 'vibration intensity discrimination', 'pressure intensity discrimination' and 'stimulated area discrimination' were represented by AR_v/RT_v, AR_p/RT_p and AR_a/RT_a, respectively.

In our preliminary test, there was no significant difference of the stimulus-evoked alpha ERD between the finger stimulation for able-bodied subjects and the same finger-projected area stimulation for amputees. Therefore, we divided the stimulated areas into four categories for the two groups of subjects (i.e. able-bodied subjects and amputees), left healthy hand (HL), right healthy hand (HR), left finger-projected area (LFP) and right finger-projected area (RFP). For each stimulation level, the size of group LFP/RFP was 4 and the size of group HL/HR was 30. We calculated alpha ERD/ERS (in the prefrontal cortex and somatosensory cortex) and ARs/RTs for each stimulated area under each level of stimulation. Then linear regression analyses were conducted between alpha ERD/ERS (contralateral somatosensory alpha ERD/ERS and prefrontal alpha ERD/ERS) and ARs/RTs. Specially, the correlations between sub-alpha ERD/ERS (lower alpha band of 8–10 Hz and upper alpha band of 10–13 Hz) and discrimination performance were also analyzed to further investigate the functional difference between sub-alpha bands.

In the experiment, the factors (independent variables) were the stimulation level (20 Hz, 30 Hz and 40 Hz in the vibration session, 150 µs, 200 µs and 250 µs in the pressure session) and stimulation area (HL, HR, LFP and RFP). There were three dependent variables, the values of alpha ERD/ERS, AR and RT. As we only focused on the brain response and discrimination performance under each sensation (vibration or pressure) rather than the difference between the two sensations, sensory modality was not considered as a factor. All the data were demonstrated normally distributed through the Kolgomorov-Smirnov test (p > 0.05) before significant analyses. A two-way ANOVA was used to statistically analyze the significant influence of the two factors (stimulation level and stimulation area) on alpha ERDs/ERSs, ARs and RTs in the two sessions, respectively. According to the two-way ANOVA, there was no interaction between the two factors, but there were respective differences in the three dependent variables. Tukey's HSD post hoc test was used to correct for multiple comparisons. For the analysis of performance-related difference, two-tailed, paired-samples t-test was used to compare the difference of alpha ERDs/ERSs between correct and incorrect discrimination.

The R² index was used to localize the discriminative information distribution among the spatial-spectral-temporal space between EEG signals in the baseline and task state. The averaged R² value distribution of all stimulation types and subjects was shown in figure 3. Discriminative information emerged out in the alpha band (see figure 3(A)) and had a well physiological localization concentrated on the channels over prefrontal and somatosensory cortices (see figure 3(B)). It was indicated that the activations of prefrontal and somatosensory cortices in the alpha frequency band were related to the tactile discrimination task.

Zoom In Zoom Out Reset image size

Based on the activated frequency band, we further analyzed the cortical activation within alpha band during the two sequential time periods in the tactile discrimination task. The grand averaged topographical distribution of cortical activation in alpha band during the two periods were illustrated in figure 4. Alpha ERD appeared over prefrontal and somatosensory cortices during the stimulation judgement period (from stimulation onset to pedal stepping). Whereas during the second period (from pedal stepping to the end of stimulation), alpha ERD was only observed in bilateral somatosensory cortices. Interestingly, the left-hemisphere topography of alpha ERD was more obvious and more posterior when compared to the right hemisphere. Analyses of equal-numbered subsets of correct and incorrect discrimination trials (see figure 5) showed similar activation patterns as figure 4. We found obvious activations of the prefrontal cortex during the stimulation judgement period no matter for the correct or incorrect discrimination. However, very subtle activations of the somatosensory cortex were observed for incorrect discrimination during the whole task. The statistical analysis indicated a significantly stronger somatosensory alpha ERD for correct discrimination (period 1 F = 10.1, p = 2.4  10⁻⁴; period 2 F = 9.8, p = 1.6  10⁻⁴) during each period when compared to that for incorrect discrimination. While there was no significant difference of prefrontal alpha ERD between the correct and incorrect discrimination during the stimulation judgement period (F = 0.93, p = 0.81).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For correct discrimination trials, ipsilateral somatosensory alpha ERS and contralateral somatosensory alpha ERD was calculated during the whole task state, while prefrontal alpha ERD was calculated only during the stimulation judgement period. Ipsilateral somatosensory alpha ERS was not found under each level of stimulation intensity for each stimulated area. For the contralateral somatosensory alpha ERD in the vibration session (see figure 6(A)), the results showed that the ERSP value at 20 Hz was approximately the same as the value at 30 Hz for each kind of stimulated area, while the value of ERSP was smaller (i.e. stronger alpha ERD) at the frequency of 40 Hz than that at 20/30 Hz. Tukey's HSD post hoc test showed that there was a statistically significant difference of ERSP value for HL between 20 and 40 Hz (F = 3.2, p = 2.1  10⁻²). When stimulated at 20 Hz and 30 Hz, the ERSP values were both statistically smaller for HR compared to that for LFP (20 Hz F = 5.1, p = 5.3  10⁻³; 30 Hz F = 5.6, p = 4.9  10⁻³). The results of contralateral somatosensory ERD in the pressure session were similar to those in the vibration session, which were shown in figure 6(B). Note that the ERSP values for HR and LFP exhibited a significant difference under the pulse width of 200 µs (F = 4.9, p = 6.1  10⁻³). Figures 6(C) and (D) showed prefrontal alpha ERD in the vibration and pressure session, respectively. We found there was no difference of ERSP values among three levels of stimulation intensities.

Zoom In Zoom Out Reset image size

Figure 7 illustrated the EEG activity over contralateral somatosensory area (channels C3 or C4) during the tactile discrimination task, for an exemplary stimulus condition (3 mA, 30 Hz, 200 µs). It was evident that a power decrease (ERD) in the alpha band (8–13 Hz) was evoked in the contralateral sensorimotor area for each stimulated area. We noticed that the activation patterns between HL and HR, LFP and RFP were similar. The alpha ERDs were stronger for HL/HR when compared to LFP/RFP. Moreover, we observed a steady-state response of power increase (ERS) around the stimulation frequency (30 Hz) throughout the task.

Zoom In Zoom Out Reset image size

AR_vs of four stimulated areas at three levels of stimulation frequencies were shown in figure 8(A). AR_vs were higher than 70% on average for both HL and HR at each level of frequency, while the results were different for LFP and RFP. For all the stimulated areas, the discrimination tasks were completed with the highest AR_vs when stimulated at 40 Hz. Tukey's HSD post hoc test indicated that AR_v for HR was statistically higher than LFP when stimulated at each frequency level (20 Hz F = 3.8, p = 8.2  10⁻³; 30 Hz F = 5.2, p = 7.4  10⁻³; 40 Hz F = 3.4, p = 1.1  10⁻²). And there was a significant difference of AR_v between HR and RFP at the frequency of 20 Hz (F = 2.8, p = 3.2  10⁻²). Figure 8(B) showed AR_ps of four stimulated areas under three levels of pulse widths. There were significant differences of AR_p between HR and LFP under the pulse widths of 150 and 250 µs (150 µs F = 3.1, p = 3  10⁻²; 250 µs F = 2.1, p = 3.9  10⁻²). AR_p was highest when stimulated with 250 µs for each stimulated area. Note that AR_vs and AR_ps in figure 8 were obviously higher than the chance level (40%). RT_v and RT_p of four stimulated area under three levels of stimulation intensities were shown in figures 9(A) and (B), respectively. Similar to the results of AR_v and AR_p, subjects spent least time to discriminate the high level of intensity when compared to the low and middle levels. Tukey's HSD post hoc test indicated that RT_v for HR was statistically shorter than LFP at the frequency of 20 Hz (F = 5.1, p = 6.3  10⁻³). For specific levels of stimulation intensity (e.g. 20, 30, and 40 Hz for AR_v, 150 and 250 µs for AR_p, 20 Hz for RT_v), the discrimination performances of healthy fingers were better than that of finger-projected areas.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Linear regression analyses were conducted between somatosensory/prefrontal alpha ERD and discrimination performance. The results indicated that there were poor correlations between the prefrontal alpha ERDs and ARs/RTs (p > 0.05), while contralateral somatosensory alpha ERDs were linearly correlated with ARs/RTs ($p \lt 0.05$). For vibration intensity discrimination, results of linear regression analyses between contralateral somatosensory alpha ERDs and AR_vs for the four stimulated areas were shown in figure 10. It was indicated that contralateral somatosensory alpha ERD was negatively correlated with AR_v. The correlation coefficient of r was −0.71 for LFP, −0.71 for RFP, −0.74 for HL and -0.73 for HR. The tendency of correlation were similar for the pressure intensity discrimination, as shown in figure 11. For LFP and RFP, r was equal to –0.71 and –0.72, respectively. While the correlation coefficients were a bit higher for HL (r = –0.75) and HR (r = –0.76). The correlation between contralateral somatosensory alpha ERD and AR_a was only analyzed for amputees (see figure 12). The ERD was negatively correlated with AR_a with r = –0.65 for LFP and r = –0.63 for RFP. Figure 13 showed the results of linear regression analyses between contralateral somatosensory alpha ERD and RT_v/RT_p for healthy finger stimulation (HL and HR). There were positively linear correlations between contralateral somatosensory alpha ERDs and RTs. For the frequency intensity discrimination, the correlation coefficient r was 0.66 and 0.69 for HL and HR, respectively. For the pressure intensity discrimination, r equaled to 0.68 for HL and 0.70 for HR. When further analyzed the correlations between contralateral somatosensory ERDs of sub-alpha bands (lower alpha band of 8–10 Hz and upper alpha band of 10–13 Hz) and discrimination performance, we found the fitting results were best for 10–13 Hz, followed by 8–13 Hz, and finally for 8–10 Hz.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

EEG studies have demonstrated the oscillatory activities over somatosensory and other task-related cortices during tactile discrimination [35, 36]. In our study, the averaged R² value distribution showed the discriminative information between baseline and task state was emerged out in the alpha band over prefrontal and somatosensory cortices (see figure 3). Alpha ERD over the somatosensory cortex was found during the whole task period, as shown in figure 4. Consistent with the result of Nikouline et al [18], alpha ERD at the primary somatosensory cortex could be evoked by median-nerve stimulation. Somatosensory alpha ERD seems to reflect the tactile information transmission and processing. Interestingly, alpha ERD over the prefrontal cortex was only observed during the stimulation judgement period in our study. With the explosion in research on prefrontal cortex functions, it has become increasingly clear that they are involved in advanced cognition, such as working memory, attention regulation and judgement [37]. During the stimulation judgement period, subjects sensed the applied stimulation and tried to judge the specific type of the current stimulus. The involvement of prefrontal cortex might contribute to the judgement and decision for the applied stimulation.

To further investigate the different roles of somatosensory and prefrontal cortices during tactile discrimination task, the cortical activation for correct and incorrect discrimination during the two sequential periods were analyzed and compared, as illustrated in figure 5. We found obvious activations of the prefrontal cortex during stimulation judgement period both for correct and incorrect discrimination with no significant difference. In contrast, the somatosensory cortical activation showed different results for correct and incorrect discrimination. As figure 5 showed, strong somatosensory alpha ERD was evoked during the two sequential periods for correct discrimination. As for incorrect discrimination, we observed weak somatosensory cortical activations during the task. The somatosensory alpha ERD during each period was significantly stronger for correct discrimination than that for incorrect discrimination. We thus got that the somatosensory alpha ERD might be related to the discrimination performance while the prefrontal alpha ERD might be not. Based on the weak somatosensory alpha ERD for incorrect discrimination, we speculated that the sensory information was not effectively transmitted along the peripheral nerve pathways to the somatosensory cortex in these incorrect trials. This speculation was consistent with the weaker somatosensory alpha ERD for amputated sides than that for healthy sides (see figure 6), as the peripheral nerve pathways might be affected after amputation. Therefore, we thought the sensory information transmitting and processing mattered the final discrimination performance, which was reflected by the somatosensory alpha ERD. Based on the strong prefrontal alpha ERD during stimulation judgement period for both correct and incorrect discrimination, it was implied to reflect the process of stimulation judgement and decision making during the tactile discrimination task no matter the final judgement was right or not. Interestingly, we observed that the left-hemisphere topography of averaged alpha ERD was more obvious when compared to the right hemisphere (see figures 4 and 5). The hand preference was suggested to be associated with an increased surface of the precentral gyrus and more intra-cortical connections in the dominant hemisphere [38]. As all subjects were right-handed in our study (see table 1), the structure-function correlation might result in a stronger stimulus-evoked activation in the left hemisphere.

For correct discrimination trials, prefrontal alpha ERD during stimulation judgement period and contralateral somatosensory alpha ERD during the whole task were calculated, respectively, as shown in figure 6. Some general conclusions and interesting phenomena were found out. For contralateral somatosensory alpha ERD, a stronger alpha ERD was observed at the frequency of 40 Hz when compared to 20/30 Hz in the vibration session (see figure 6(A)), which was similar in the pressure session (see figure 6(B)). Increasing the stimulation frequency resulted in an increase in the firing rate of activated neurons. While increasing the pulse width leaded to the recruitment of additional neurons [39–41]. The stimulation frequency and pulse width had systematic and cooperative effects on perceived tactile intensity which depended on the total stimulation charge applied per second [8]. We found no significant difference of alpha ERD between 30 Hz and 20 Hz . It might due to the non-linear relationship between the stimulation parameters and stimulus-evoked alpha oscillations, which deserved to be further investigated in the future. In contrast, there was little difference of prefrontal alpha ERD among three levels of stimulation intensities, as shown in figures 6(C) and (D). It was implied that the value of prefrontal alpha ERD was independent of stimulation parameters. We thus speculated that somatosensory alpha ERD was evoked by the tactile stimulation itself, which might be related to the functionality of the upscending sensory pathway and sensory information processing. Prefrontal alpha ERD was suggested to be induced by the stimulation judgement, which might represent the process of advanced information integration and decision making. These were in accordance with the results of different activated cortical areas during the two sequential time periods of the task (see figure 4).

In order to explore the functional role of alpha ERDs (somatosensory alpha ERD and prefrontal alpha ERD) during the tactile discrimination task, we calculated the correlation coefficients between alpha ERDs and subjects' discrimination performance. Only the somatosensory alpha ERD was found to be correlated with subjects' discrimination performance. The contralateral somatosensory alpha ERDs showed negative correlations with ARs and positive correlations with RTs in the linear regression analyses (refer to figures 10, 11, 12 and 13). We found the three shapes of fitting points (representing three levels of stimulation intensities) were uniformly distributed around the regression line, indicating a generalized correlation for different stimulation parameters. The linear correlations between contralateral somatosensory alpha ERD and discrimination performance were consistent with our speculations about the functional role of somatosensory cortex in facilitating tactile information processing. In contrast, the prefrontal cortex was not associated to the discrimination performance. And its functional role of reflecting the process of stimulation judgement and decision making was only recruited during the stimulation judgement period. Furthermore, comparing the correlations between somatosensory ERDs of sub-alpha bands (8–10 Hz and 10–13 Hz) and discrimination performance, we found the fitting results were better for 10–13 Hz than 8–10 Hz. It was suggested that the functional roles of the two sub-alpha bands also differed in the current tactile discrimination task.

Alpha rhythm oscillations have been suggested to act as a gateway from sensation to cognition. Particularly, the amplitude and phase of the alpha rhythm entails a cortical mechanism which paces the access of sensory information into the cognitive system [42]. Evidence supporting this idea includes correlations between the amplitude/phase of alpha rhythm oscillations and behavioral performance in perception [43–45], spatial attention [46, 47] and working memory [48, 49]. Top-down mediated modulation of the alpha activity in somatosensory cortices could facilitate the processing of a specific input stream through an increase of excitability of task-relevant regions [50]. For example, Kelly et al found a relationship between alpha-lateralization and visual target discriminability/reaction time [51]. Moreover, recent studies have found that alpha activity appears to play a role in participants' subjective confidence in their visual judgements [52]. In general, the correlations between the alpha rhythm oscillations and behavioural performance were mostly investigated for visual or auditory processing, whereas few studies focused on tactile processing. In fact, bilateral alpha ERD has been observed during somatosensory stimulation, e.g. electrical stimulation of the median nerve [18] or tactile stimulation of the index finger [53, 54]. The current study investigated the correlations between electrotactile stimulation-evoked alpha ERDs and tactile discrimination performance for both amputees and able-bodied subjects, which might reveal more physiological mechanisms underlying the alpha rhythm oscillations for tactile processing.

The connection of EEG oscillations and tactile acuity was shown here by way of significantly linear correlations between contralateral somatosensory alpha ERD and tactile discrimination performance. Given these correlations, assessments of contralateral somatosensory alpha ERD during the sensory rehabilitation of the amputated limb could be used as an objective method for the tactile acuity evaluation. There are several clinical values of having EEG-derived bio-markers. Firstly, EEG is easily applicable in the clinical setting. Specifically, this non-invasive imaging tool is acceptable for amputees. Meanwhile, the EEG-based markers could indicate effects of therapeutic interventions and the course of treatment-induced recovery for amputees. Moreover, in terms of sensory feedback of the myoelectric prosthetic hand, the quality of evoked physiological sensation could be assessed using the index of somatosensory alpha ERD. Based on the correlations between contralateral somatosensory alpha ERD and ARs, the stimulation parameter with superior evoked-ERD could be selected to insure a better tactile discrimination performance, which could contribute to a more accurate transmission of electrotactile stimulation encode.

Several limitations of the current study should be mentioned. In our experiment, we only recruited four amputees, which may be not applicable for the statistical analysis of two-way ANOVA. Meanwhile, the group size of amputees was not sufficient enough for the linear regression analyses. More amputees should be recruited in the future to get a more reliable statistical analyses and a better fitting effectiveness. In addition, the effectiveness of electrical stimulation in producing natural pressure and vibration sensations was not assessed in the current study. A future work is necessary to compare the sensations evoked by TENS with natural sensations evoked by pressure or vibration instrument in order to further advancing this technique in realistic applications. Furthermore, the four recruited amputees all reported phantom limb pain, and this pain might have an effect on the somatosensory alpha ERD. However, the phantom limb pain-induced effect should be equivalent to all stimulation conditions, which is unlikely to be a confounding factor for this study. Separating the effect of phantom limb pain from tactile sensation would be an interesting study and is worthy of future investigation.