Human subject recruitment
All experiments were performed adhering to relevant guidelines and regulations, in accordance with the procedure described in the protocol approved by Institutional Review Board, Texas A&M University (IRB2018-1583D). Eight healthy human subjects in age 22–35 (with average of 26.1), one female and 7 males participated in the study. All subjects were right-handed. Subjects with neurological disorder, cognitive impairment, upper limb deformity, and any known allergic problem to skin adhesive were excluded from the study. All subjects provided their informed consent for the experimentation according to the approved IRB protocol.
System implementation
A biphasic voltage-controlled electrical stimulator was designed for providing transcutaneous electrical stimulation to subjects. Voltage-controlled stimulator provides an effective and easy way to generate electrical stimuli. The system consisted of a microcontroller (Particle Photon with STM32 ARM Cortex M3) to generate input pulse-width-modulated (PWM) waveforms followed by a level shifter to increase the voltage level and convert it into biphasic square-wave output. Microcontroller was programmed to generate biphasic square-wave output with frequency ranging over 0–10 kHz and duty factor over 0–100%, according to the operator input. A small signal NPN transistor (MMBT3904, On Semiconductor, AZ) was used for voltage level shifting to generate the actual voltage stimulus ranging over 3–30 V. The biphasic voltage stimuli were generated by an H-bridge circuit, composed of CMOS n-channel and p-channel FET (field effect transistor) pairs (CD4007UE, Texas Instrument, TX) [48]. The system was powered by a rechargeable Li-Po battery. A step-up voltage converter was used for converting 3–4 V from Li-Po battery to high voltage levels up to 30 V.
Custom designed transcutaneous gel electrodes were used to deliver the electrical stimulus to the skin over the target muscle. The custom designed electrodes were made using the reusable self-adhesive electrode. These reusable electrodes were customized in size and multithreaded connecting wires were stacked on top using silver conductive epoxy [23, 48]. Electrodes were customized with a small footprint (approx. 1.2 × 0.8 cm2), to ensure high localization of the electrical stimulation and identify the appropriate electrode locations with maximum effect. The self-adhesive hydrogel was pasted on the electrodes and additional latex-free adhesive tape was taped over the electrodes, to ensure stable contact of electrodes onto the skin during arm movements.
Two gyroscope sensors (MPU9250) were strapped onto both left and right forearms, one for each arm, to record the angular data for both elbow joints. An elastic strap with Velcro was used to fasten the gyroscope sensors on subjects’ arms to allow for minimum perturbations. The sensitivity scale factor of 131 (LSB)/°/s and a full-scale range of ± 250°/s is used for the selected gyroscope sensors. The gyroscope data was digitized using a built-in 16-bit ADC in MPU9250 to provide high-resolution data, and then sampled at 10 Hz by the microcontroller. Gyroscopes were calibrated every time before the experiment, for data integrity. The gyroscope data was delivered to the microcontroller via SPI interface and saved to the computer via USB interface. As gyroscope provides the derivative of an angle, the desired angle values were calculated by integrating the gyroscope data over time.
Experiment procedure
Parameter selection for electrical stimulus
For the all experiments, we selected biphasic square-wave electrical stimulus for charge balancing. 50% duty factor and 10-ms inter-pulse interval (i.e., 100 Hz) were also selected as default stimulation parameters, based on previous successful experiments that showed the effect of electrical stimulation on proprioceptive modulation [28, 47]. The amplitude range of the stimulation was determined per subject based on subjective perception, between perception and discomfort thresholds. As the stimulation was continuously applied during the given “stimulation-on” duration, determined by each experimental condition, we didn’t specify the stimulus train duration and Inter-stimulus interval.
Experiment I: identification of electrode placement
The first experiment was designed to identify the location of electrodes for transcutaneous electrical stimulation. We selected four different locations of bipolar electrodes, on biceps brachii short head and brachioradialis. Spindle afferent feedback from these two muscles, across the elbow joint, contributes to the perception of the elbow joint angle [50]. As it is hard to determine the accessibility of transcutaneous current to reach the muscle spindle via the muscle belly and the myotendinous junction areas, we targeted both of those areas for two synergistic elbow flexor muscles. Figure 2 shows all four identified electrode locations and combination of these four locations resulted in 10 different locations for a pair of electrodes.
With a pair of electrodes placed on selected locations, we set the stimulation voltage amplitude as 5 V. No subject reported any sensation at 5 V stimulation voltage. The voltage was then gradually increased and the voltage amplitude at which subjects first reported some paresthesia or any stimulation-based sensation was recorded as perception threshold. The voltage amplitude was further increased to a level until subjects reported any discomfort. We set the stimulation voltage right below this discomfort level where subjects were comfortable to be stimulated for long duration. Once the required voltage for proprioceptive illusion was identified, the electrical stimulation was turned on and subjects were asked to report the subjective feeling of arm flexion or extension (i.e., proprioceptive illusion) by 1–5 subjective rating while providing stimulation for 5-s duration. This was repeated two times in series. The 1–5 scale was defined to evaluate the intensity of proprioceptive illusion, as 1 represents no illusion and 5 represents very strong illusion (1—nothing, 2—barely perceivable, 3—perceivable, 4—strong, 5—very strong).
Stimulation frequency was fixed at 100 Hz for the Exp. I and the duty factor of the biphasic stimulus was fixed as 50%, according to the previously successful parameters for evoking electrotactile feedback [51]. Subjects were asked to maintain elbow joint angle at 90˚, where 180˚ means fully extended elbow joint. The selected voltage level and the corresponding subjective rating of proprioceptive illusion were recorded. Based on this subjective rating, we intended to select the best electrode pair location to be used for the following experiments to test proprioceptive illusion. Note that the experiment I was executed only for the first three subjects. It is because the location for the strongest proprioceptive illusion was clear and consistent for the first three subjects. To minimize potential aftereffect of the stimulation, experiments for the other five subjects used the selected electrode location based on the result of the first three subjects.
Experiment II: characterization of electrical stimulation parameters
The second experiment was designed to identify the best amplitude and frequency of electrical stimulation for the maximal proprioceptive illusion, with positioning electrodes on the best location found at the first experiment. Subjects were asked to place their arm at rest with their elbow firmly placed onto the 90° armrest for reference [21]. Although the armrest stays at the location during the experiment to guide the joint angle of the right arm, subjects were asked to actively maintain the guided elbow joint angle and not to rest on the armrest. In other words, subjects were barely touching it instead of resting on it. Subjects were then blindfolded, and voltage level was gradually increased. Subjects were asked to report when the electrotactile feedback, generally described as tingling, started to set the perception threshold (Vth). The voltage level was further increased, until the subject reported any discomfort. The maximum stimulation voltage right below the discomfort range was set as the maximum voltage level (Vmax). The amplitude of the stimulation used for the experiment was determined per subject based on subjects’ report of the maximal proprioceptive illusion, by slowly increasing the amplitude from Vth to Vmax. Note that, standard psychophysics methods were not used for measurement of the thresholds and the amplitude of the stimulation used, which is a limitation of this study.
We also identified the appropriate frequency of electrical stimulation for the proprioceptive modulation, with the same procedure of arm resting and blindfold. For this part, the voltage was fixed to the Vmax found earlier with the frequency fixed at 100 Hz. Subjects were asked to report the effect when the stimulation frequency was changed from 100 Hz. Subjects were provided stimulation with a set of frequencies of 30, 100, 300, 1000, and 3000 (Hz) and asked to rate the proprioceptive illusion from 1–5 for each frequency.
Experiment III: arm matching experiment
The third experiment was designed to quantify the angular displacement induced by transcutaneous electrical stimulation. Arm-matching test was selected for quantification of the illusory flexion/extension of the elbow joint, as it has been proved as a reliable way to evaluate proprioceptive illusion in prior works [1, 4, 5]. As shown in Fig. 3, subjects were asked to place their arm at rest with their elbow firmly placed onto the 90° or 135° armrest for reference. Subjects were clearly instructed to actively maintain the guided elbow joint angle and not to rest on the armrest. To ensure consistent muscle pre-conditioning, the subjects were instructed to bring their right arm close to shoulder before maintaining it at reference angle at start of each trial. Note that, such muscle conditioning/thixotropy maximizes the illusory effect of elbow extension, as shown in previous vibration-based proprioceptive modulation studies [1, 33].
Subjects were also asked to relax their arm muscles and keep their right arm stationary. Subjects were blindfolded during the experiment to avoid any bias from the visual feedback. Each session was composed of baseline measurement, stimulation test, and aftereffect test. First, for baseline measure, subjects were asked to do arm matching without any stimulation provided. Initially, subjects were asked to keep their right arm stationary at specified reference angle and keep their left arm in fully extended posture. When they hear the audio command, “Flex”, they were asked to use their left arm to match the right arm elbow joint angle. When they hear the audio command “Go back”, they were asked to return their left arm to the fully extended posture resting their arm on the desk. A constant time delay of 4 s was provided between each audio command for subjects to complete the required arm movement and keep the elbow joint angle at that angle. This arm-matching test was repeated for four times in series to ensure that the baseline value was consistent. The gyroscope data for this arm matching sequence was saved to a computer. Second, to measure the effect of electrical stimulation, electrical stimulation was applied on the best identified location with best identified parameters. Subjects were then asked to move their left arm to match the elbow joint angle between the left and right elbow joint. According to the audio commands, the arm-matching test was repeated for two times in series. Third, to measure the aftereffect of electrical stimulation, the arm-matching test was repeated for two times in series, as shown in Fig. 4. The gyroscope data for stimulation on/off sequences was saved to a computer. The same procedure (2 consecutive sessions) was conducted for two different reference joint angles: 90° and 135°, with a minimum of 1-min interval between sessions for each reference joint angle. The session order was counterbalanced among subjects with randomization (in a random order, 4 subjects conducted 90° first and the other 4 subjects conducted 135° first). In summary, two sessions for two different reference target elbow joint angles were conducted for all the subjects following the above-mentioned procedure. Within the single session, arm matching test was repeated for four times to measure a baseline, repeated for another two times with stimulation applied, and repeated for another two times after stimulation was turned off (see Fig. 3b).
At the end of the arm matching experiment, we collected subjective descriptions about the proprioceptive illusion, from all eight subjects, to get more insight about the principle of the proprioceptive illusion. In addition, we also collected subjective description of the accompanying unnatural sensation for all eight subjects as we know that transcutaneous electrical stimulation often evokes unnatural sensation, referred to as tingling or paresthesia.
Experiment IV: Pinocchio illusion experiment
The fourth experiment was designed to confirm that the induced illusion is proprioceptive in nature. Pinocchio illusion experiment has been used in past [52, 53], to establish and understand the proprioceptive illusion. This classical experiment involves the subject touching their nose with a fingertip of the same arm on which the stimulus for inducing proprioceptive illusion is applied. Subjects were blindfolded for this experiment and we used the best electrode location and stimulation parameters identified in Exps. I and II, to induce the maximum proprioceptive illusion while minimizing potential bias caused by visual feedback. Based on the direction of illusion, subjects perceived their nose tip elongated or shrunk.
Subjects were instructed to touch their nose tip using their right index fingertip after being blindfolded. The biphasic electrical stimulation was provided for a duration of 10 s on their right arm with electrodes placed on previously identified locations. Subjects were asked to keep their index fingertip on the nose tip for 10 s after the stimulation was turned off, to report any aftereffect. Subjects were asked to select the pictorial representation of nose (on a scale of 1–5 for both elongation and shrinkage) from a series of nose representations that best describes their feeling (as shown in Fig. 5).
Data analysis and statistics
For the subjects’ data in the arm-matching test, elbow angles of the left arm were calculated as the average value within effective duration of each measure. To define the effective duration, we first defined both the start and end of the arm-matching period, as the times when the elbow angle is crossing the value 30° larger than the target elbow angle (165° for 135° target and 120° for 90° target), as shown in Fig. 4. We then excluded 1.5 s from the start of the arm-matching period, to minimize the transient response. We additionally excluded 0.5 s at the end of the arm-matching period, as shown in Fig. 4, to minimize the effect of undershoot before movement. As a result, the effective duration for each measure is around 2 s, and the average of data for this effective duration was calculated and used. As shown in Fig. 4, in each subject, the average of the first four arm-matching periods (#1 to #4) was used as a baseline value before applying stimulation, the average of the next two measures (#5, #6) was used as a value for evaluating stimulation effect, and the average of the last two measures (#7, #8) was used as a value for evaluating stimulation aftereffect.
For statistical analysis of the data, one-way repeated ANOVA test was performed at the 95% confidence level, to conduct a robust statistical test not to be falsified by any small difference in variances. We selected one-way repeated ANOVA with having just time (before, during, and after stimulation) as an independent variable, instead of two-way ANOVA with having time and the reference elbow joint angle as two independent variables. It is because different reference elbow joint angles may set different conditions of muscle spindle and the result for one reference joint angle may be under- or overrepresented by the result for the other reference joint angle. To verify that the data satisfies the prerequisites for the ANOVA test, we tested normality of data distribution using the Kolmogorov–Smirnov test of normality. All datasets satisfied the condition of p > 0.05 and normality could be assumed. We also applied Bonferroni correction, and then used p < 0.05 as the condition for statistical significance. IBM SPSS Statistics was used as a statistical software.