Multi Surface Electrodes Nerve Bundles Stimulation on the Wrist: Modified Location of Tactile Sensation on the Palm

Transcutaneous electrical nerve stimulation (TENS) of nerve bundles, such as ulnar and median nerves, is a technique used to evoke tactile sensation on the hand. Considering this technique does not require electrodes to be attached to the hand, it can induce sensation without interrupting the interaction between the hand and environment. Although this technology has most commonly been explored for restoring sensory feedback in amputees, it can also be applied to tactile devices for non-disabled individuals to solve problems that recent tactile interfaces have been facing such as interference of the device with the interaction between the hand and environment. We propose TENS at the wrist with multi-electrode targeting the cutaneous nerve bundles as a method to induce tactile sensation with higher spacial resolution, and without unintended tactile sensation at the upper arm. In this study, we demonstrated through psychophysical experiments and finite element simulation that our method induces sensation only at the hand, and that it could control the location of induced tactile sensation in the circumference direction.


I. INTRODUCTION
In the field of virtual reality (VR) and human-computer interaction (HCI), several technologies for tactile display have been developed. There are numerous studies on tactile display, and several related applications have been proposed. Iwata et al. developed a tactile display using a pin array with a kinematic mechanism [1]. Yoshida et al. proposed a handheld haptic display using a similar method [2]. Minamizawa et al. developed GravityGrabber, which can present the vertical and shearing force to the surface of the finger by actuating motors and belts [3]. Tactile displays are also applied to teleexistance (telepresence), where a humanoid robot is placed at a remote location [4], and the user who wears the tactile display device The associate editor coordinating the review of this manuscript and approving it for publication was Arianna Dulizia . perceives tactile sensation when the robot interacts with an object in a remote location. In these studies, the tactile system requires the user to perceive the same sensation as they are in the remote or virtual environment. Therefore, the area at which the tactile sensation occurs were allowed to be covered by the tactile display.
However, in tactile interfaces for applications such as Augmented Reality and Mixed Reality (AR/MR), the device should not cover the sensory presentation position. In a typical AR or MR technology, the user visually recognizes virtual objects and textures superimposed over the view of the real world. This concept should also be applied to tactile sensory information. Specifically, haptic and tactile AR technologies must be able to modify the sensation when the users touch, grab, or trace real objects. With this background, various tactile technologies that can be applied in AR have been proposed. Some approaches stimulate a distant location to create tactile illusion using physical stimulation such as vibration [5] and ultrasound [6], while others use transcutaneous electrical stimulation (TENS) against the finger [7], [8], [9]. While these technologies create or modify tactile sensation minimizing the interruption of the interaction between the hand and environment, tactile induction is limited to a part of the hand, and the device needs to be attached at other positions of the hand, which inhibits hand movement. These issues are critical in AR or MR where real-world exploration is as important as interaction with the virtual environment.
On the other hand, TENS against the nerve bundles at the arm, such as median and ulnar nerves, is being studied in the field of medicine. Because this method can present tactile sensation at the hand with electrodes attached at the forearm or upper arm, it has been studied to cure phantom pain [10] and obtain tactile feedback for prosthetic hand users [11], [12], [13], [14], [15], [16]. TENS for tactile sensation in the hand is obtained by targeting the nerves that innervates cutaneous sensation. Cutaneous sensation in the hand is innervated by the ulnar, median, and radial nerves, all of which connect to the central nervous system through the arm. The ulnar nerve innervates the area on the little finger, and the radial nerve innervates the area from the outside of the thumb to the back half of the ring finger. The median nerve innervates the rest of the hand (Fig. 1). Thus, tactile induction with TENS at the nerve bundles does not require the hand to be covered, which has the potential to become a tactile technology for AR or MR.
However, this advantage of the method inherits a major issue, which is unclear spacial resolution of the induced sensation. Previous studies revealed that TENS targeting the ulnar and median nerves selectively each induce sensation to the area innervated by each nerve for non-disabled subjects [12], [14]. It is also known that using an array of small electrodes along the upper arm induce sensation at a more detailed area within each nerve innervation area, although the pattern between the stimulation position and the location of induced sensation was not consistent between subjects [13]. In order for this method to be effectively integrated with AR and MR technologies, sensation induction at higher spacial resolution in a general way is necessary. In addition, unintended tactile sensation was reported at the forearm and upper arm in some studies [13], [14], which should be removed for natural experience as well.
We propose TENS at the wrist with multi-electrode as a method to induce tactile sensation: 1. at the hand without covering it, 2. with higher spacial resolution than that for each innervated area, and 3. without unintended tactile sensation at the forearm or upper arm.
Nerve bundles, such as the ulnar, median, and radial nerves, transmit somatosensory information, such as the sense of vibration and pressure, which are then perceived at the mechanical receptors such as the Markel cell, Meissner corpuscle, and Pacini corpuscle [17]. Considering that the nerve fibers connected to these cells are organized into the nerve bundles [18], the area innervated by the nerve bundle will be wider as it gets closer to the trunk. Therefore, we hypothesized that stimulating the nerve bundle using electrode positions closer to the hand would reduce the tactile sensation induced at the unintended areas. Moreover, the area where the tactile sensation is induced apparently changes within the innervation area of the targeted nerve bundle depending on which part of the nerve bundle is stimulated. Accordingly, we hypothesized that the position of tactile sensation induced by the nerve bundle stimulation changes depending on the current density distribution on the nerve bundles.
The contribution of this work is that it introduces multiple-electrode TENS at the wrist to the field of AR and MR tactile technology and challenged the issue of low spacial resolution through psycophysical experiments and simulation by concluding that it can control the location of induced tactile sensation in the hand without covering it. We first tested the most effective position for stimulation between the wrist, elbow, and upper arm, where the nerves run close to the skin surface. As hypothesized, stimulation at the wrist induced sensation only in the hand, while stimulation at the other positions induced sensation in the arm. Then, a finite element simulation was conducted to show that the current density distribution formed by multi-electrode wrist stimulation on the nerve bundles differed depending on the position of stimulation. Lastly, we conducted a psychophysical experiment based on the simulation to demonstrate that the area of tactile position changed with the position of the electrode. The results showed that it is possible to modify the position of tactile sensation in the circumference direction of the wrist, which implies that the current density distribution within the nerve bundles varied with the stimulation position for real subjects.

A. TACTILE TECHNOLOGIES FOR AR AND MR
Various tactile induction technologies that can be applied in AR or MR have been proposed (Table 1 ). One idea proposed VOLUME 11, 2023 by Teng et al. is to control the timing of tactile induction [19]. They developed Touch&Fold, where the mechanical component presents several types of haptic effects when they interact with virtual objects and quickly tucks away when the user interacts with real-world objects. Other methods make use of the mechanisms of human perception and achieve both artificial sensation and interaction with real environment simultaneously. These methods can primarily be divided into two approaches.
The first approach is to mechanically stimulate a distant location and create a tactile illusion. Ando et al. developed an example of tactile AR display technology [5] called SmartFinger. This technology creates an illusion of unevenness and texture on the surface of an object being traced by presenting vibrations from an oscillator attached to the fingernail of the finger tracing the object. SkinHaptics, which was presented by Spelmezan et al. [6] uses ultrasound stimulation at the back of the hand to induce a slight tactile sensation on the palm. They also showed that ultrasound stimulation at the palm induces sensation on the back of the hand. Tao et al. developed a wearable device to modify the softness sensation of real objects by restricting finger pad deformation using a hollow frame [20].
The second approach is to use TENS. Because TENS only requires a light weight, small, and reasonable device, many applications with TENS are being proposed in the field of VR and HCI. TENS is a technology often used for inducing various sensations such as taste [21], vestibular sensation [22], intranasal chemical sensation [23] and muscle contraction [24] through the epidermis along the nerves. Tactile sensation is no exception and is achieved by targeting the cutaneous nerves and mechanical receptors under the skin. Kajimoto et al. used a thin electrode matrix array and induced tactile sensation by electrically stimulating the mechanical receptors and adjacent nerve fibers [7]. Their SmartTouch embeds a sensor to detect contact with objects and modifies the tactile information of a real object. Miyamoto et al. combined this technology and electrical muscle stimulation (EMS) to reproduce interaction with an object in midair [8]. Yoshimoto et al. developed a TENS method targeting the finger pad of the index finger and thumb by stimulating the second joint of each finger, to modify roughness perception of a real object [9]. They demonstrated that both fine-and macro-roughness perception enhanced with TENS, and that the induced sensation fused well with the real material particularly with lower frequency stimulation.
In general, these tactile displays comprise of mechanical components that should be attached to the opposite side of the sensory presentation such as the fingernail or back of the hand. Even when the artificial tactile induction targets a specific location, this impairs natural real-world exploration using the hand, such as touch, rub, or grab. This concept of impairment is considered one of the four important factors for wearable system haptic systems [25]. This also makes it difficult to present sensation at other positions of the entire hand, including the back of the hand.

B. TENS AGAINST THE ARM NERVE BUNDLES
TENS against the nerve bundles at the arm is known to induce sensation at the hand and is being studied as a method to obtain tactile feedback for prosthetic hand users. Chai et al. applied TENS against both amputees and non-disabled subjects and reported that similar sensory modalities of touch, pressure, buzz, vibration, and numb were evoked in both groups [11]. The mapping of the stimulation position and the sensory location for amputees were stable for a long term, which indicates that re-mapping every time in wearing a prosthetic hand is unnecessary when obtaining feedback. Forst et al. used TENS targeting the ulnar and median nerves at the elbow for non-disabled subjects and measured the voltage amplitude and pulse duration thresholds to examine threshold characteristics [12]. They reported that both median and ulnar nerve stimulation induced sensation at each of their respective innervation areas at the hand. The elicited sensation was described as paraesthesia-like sensation or occasional natural sensation.
Some studies have used this method to induce more complex tactile patterns. Shin et al. manipulated the stimulation waveform to reproduce three types of tactile experience, which were single tap, press-and-hold, and double tap, and reported that subjects were able to discriminate between them. Scarpelli et al. achieved to evoke an apparent movement sensation, which is a sensation that moves within the palm [14]. This is based on a tactile phenomenon where differences in the intensity of two tactile stimuli at different locations affect the location of perceived sensation. By modifying the intensity of stimulation to the median and ulnar nerves dynamically, the subjects were able to perceive the movement sensation. Lastly, Vargas et al. applied the TENS for prosthetic users and demonstrated that the tactile feedback enhance the object shape and surface topology recognition [15].
The advantage of not having to cover the hand is necessarily accompanied by the issue of the spacial resolution of the tactile induction. Because the position of stimulation is distant from the location of sensation, investigation on the relationship between them is necessary. In fact, most of the above studies starts with the mapping of the stimulation position and sensory location. For example, Shin et al. [13] and Vargas et al. [15] used 16 electrodes arranged in two rows along the upper arm and evoked sensation in fine spacial resolution such as at the finger pad of some fingers. However, they report that there was no consistent pattern in the mapping among subjects because of large individual difference in the body structure at the upper arm. This means that this mapping has to be created for each individual from the beginning. Considering that prosthetic hands are worn on a daily basis, this is not considered a major issue because once this mapping is created, it is known to be stable for a long term [11]. On the other hand, in terms of tactile technology for AR and MR where easy adaptation is important, a general method to control the sensory location is necessary. The method that is currently identified to be consistent in sensory location is TABLE 1. AR and MR tactile technologies discussed in Section II-A. Generally, the target location and stimulation position is different. However, the hand movement is impaired because the stimulation involves mechanical components at the hand.
that of Forst et al. [12] and Scarpelli et al. [14] where they targeted the median and ulnar nerve selectively and confirmed that the sensation was induced at regions innervated by each nerve. This current spacial resolution should apparently be enhanced to increase the range of tactile expression. Additionally, TENS at the upper arm is known to elicit unintended sensation at the forearm and upper arm [13], [14], which also needs to be addressed to achieve a natural tactile experience.
We aim to induce tactile sensation only at the hand, and to the entire hand in a higher spacial resolution than conventional TENS method by applying multi-electrode stimulation along the wrist. To the best of our knowledge, this study is the first to develop such tactile technology which could be integrated with AR or MR experiences.

III. FIRST EXPERIMENT: THE EFFECT OF ELECTRODE POSITION ON THE TACTILE SENSATION INDUCED BY NERVE BUNDLE STIMULATION A. METHOD OF THE FIRST EXPERIMENT
To resolve the issue that the conventional nerve bundle stimulation induces tactile sensation at the unintended area, we investigated the effect of the position of the electrode on the area of induced tactile sensation.
Eleven healthy adults (11 males) participated in the experiment. The study protocol was approved by the local ethics research committee at The University of Tokyo , and the participants signed the letter of consent after being explained about the experiment. The study protocol was performed in accordance with the ethical standards outlined in the Declaration of Helsinki.

1) STIMULATION POSITIONS
Six positions were used for stimulation along the median and ulnar nerves for the wrist, elbow, and upper arm, respectively (Fig. 2). These positions were used as the positions where the nerves run close to the epidermis based on anatomical findings. A cathodal gel electrode (Vitrode F-150S, Nihon Kohden Corporation) was used for the stimulation electrode because it is clear that cathodal stimulation evokes nerve activity that runs parallel to the skin [26]. An anodal gel electrode (Vitrode F-150S, Nihon Kohden Corporation) was fixed under the ulnar styloid process because the nerve does not run at this area. Each electrode was named as E-MU, E-ME, E-MW, E-UU, E-UE, and E-UW. The positions of E-MU,  E-ME, and E-MW target the median nerve, whereas that of E-UU, E-UE, and E-UW targets the ulnar nerve. To minimize the effect of individual differences in nerve locations, the cathodes were placed according to our anatomical landmarks shown in Table 2. The subjects were asked to describe the area of tactile sensation for each of the six-electrode position conditions.

2) STIMULATION CONFIGURATION
Constant current was used as the waveform of the electrical stimulation. Preliminary trials revealed that periodic waveform, which is often used for TENS, induces sensation around the anodes at stronger intensity, although they were not placed directly above the nerve. Since our main objective was to investigate the location of induced sensation, constant current was adopted to focus on the relationship between the targeted nerve and the location of induced sensation. In addition, to suppress the instantaneous strong sensation, the stimulation current value slowly approached to the target value. VOLUME 11, 2023 Stimulation at each position was conducted at five different current values to investigate the effect of the intensity of stimulation. Because subjects are assumed to have different sensitivity against electrical stimulation, the current value relied on the threshold current value to control the subjective intensity between subjects. Based on preliminary observations, the five current values were determined as T , 1.
× T , and maximum current strength, where T is the threshold current value, at which tactile sensation occurs. The maximum current strength was the condition at the maximum intensity which does not cause discomfort to the subject in the range of 5 mA or less. Note that when the calculated current value was over 5 mA, the stimulation was conducted at 5 mA for safety reasons, and the collected data was excluded from the analysis.

3) PROTOCOL
Stimulation was performed at each position according to a randomly determined order. First, for each position condition, the threshold current was measured by parameter estimation through sequential testing (PEST method) [12], [27], [28]. In this method, the subjects were repeatedly asked to answer whether or not they felt tactile sensation in their hand in response to a stimulation presented according to a specific procedure, where the current of stimulation increased if the subject reported no sensation and vice versa. During the threshold determination, each stimulation was conducted for 1, 000 ms, and the initial current and step size were 0.25 mA and 0.375 mA, respectively. The step size decreased by a factor of 0.7 each time the subject's response differed from the previous response until this occurred three times. These values were determined based on preliminary trials by the experimenter so that the threshold measurement is conducted within the time constraints of the experiment.
Then, the stimulation was conducted at five different intensities in ascending order. During each stimulation, the subjects had to describe the areas of tactile sensation and maximum point of tactile sensation that they felt using an original drawing software shown in Fig. 3, developed in Unity 2019.3.14f1. In this software, a three-dimensional view from the left upper arm to the fingertips and the cathodal electrode position was pre-drawn. The pre-drawn view was internally distinguished into two parts, the hand and the arm, where the arm indicated the whole area except for the hand (Fig. 4). The top view of each part comprised 67,177 pixels and 235,510 pixels, respectively. There were two drawing modes, the area mode, and the maximum point of sensation mode. In the area mode, the subjects were to draw the area of sensation, and they could draw, undo, redo, and change the thickness of the drawing line. In the maximum point of sensation mode, the subjects were to plot the maximum point of sensation in all three views. Both the vertical and horizontal coordinates of the point synchronized within the three views to ensure there was no geometric contradiction. The software was run on Surface pro (Microsoft Inc.), and the subjects drew the area  The stimulation waveform was a trapezoidal wave that linearly increased up to the stimulation current and remained constant for 2, 000 ms. All stimulation combinations were conducted once, and 30 stimulations were conducted in total (six electrode positions ×5 current strength conditions).
After the recording was done for all of the conditions, participants responded to a questionnaire in which they verbalized the sensation they perceived during the experiment. show the size of the areas of tactile sensation on the hand and rest of the arm. We applied an aligned rank transform (ART) [29] to the size of the area of tactile sensation on the arm and hand to analyze the interaction effect with nonparametric data. Then, we conducted a statistical analysis of the variance (ANOVA) on the aligned ranks for the area of tactile sensation for both the arm and hand. The results for the arm showed significant effect in the electrode position factor (F(5, 50) = 63.52, p < 0.001, η 2 p = 0.23), current value factor (F(4, 40) = 23.72, p < 0.001,

C. DISCUSSION OF THE FIRST EXPERIMENT
All of the subjects were able to perceive sensation at locations other than directly under the electrode in all position conditions. According to the final questionnaire, most of the subjects described the induced sensation as either vibration or paranesthesia-like sensation which is consistent with which was reported in previous work [12], [13], [15]. It was also observed that several subjects used various onomatopoeias to describe the sensation, which means that it was difficult to match the sensation with natural experience. Fig. 5, Figs. 6a, and b show that the area of tactile sensation induced by the electrical nerve stimulation increased as the stimulus intensity increased. Additionally, in conditions where the wrist was stimulated, the area of tactile sensation on the arm was smaller compared to the other conditions. Fig. 5, Figs. 6c, d, and e show that median nerve stimulations (E-MU, E-ME, E-MW cathodal stimulation) and ulnar nerve stimulations (E-UU, E-UE, E-UW cathodal stimulation) induced tactile sensation on the thumb side of the hand and little finger side, respectively. Furthermore, the maximum sensation points of the y direction in both the wrist stimulation were the highest in all conditions, while these become smaller as the stimulation position becomes more proximal. This is because the nerve bundles innervating tactile information in the forearm and upper arm are also stimulated. Although it is known that these nerves branch at the brachial plexus are located around the clavicle, it became more difficult to stimulate only the median and ulnar nerves as the stimulation position gets closer to the center of the body. In addition, some subject felt stronger sensation at the forearm and upper arm than at the hand, which implies that stronger current distribution was created at those nerve bundles. Therefore, the maximum point of sensation was recorded at the forearm for those subjects, while others recorded at the hand. This trend was remarkable especially in the upper arm conditions, where the the effect of individual differences in muscle and nerve placement was also large resulting in large variance.
Results of this experiment show that the area of tactile sensation in the forearm and upper arm in wrist stimulation is smaller than the other cathodal position conditions. This indicates that the wrist is suitable as the position of stimulation for tactile display presenting tactile sensation on the hand, considering the sensation induced in the unintended area (arm) can be reduced compared to stimulating other positions. Additionally, although the points of the maximum tactile sensation in the elbow and upper arm stimulation were at the forearm or near the wrist, these points in the wrist stimulation were located on the hand.
Regarding the stimulation strength, Figs. 5 and 6 show that the higher current induces a larger area of tactile sensation, although the positions of the maximum tactile sensation were similar regardless of the stimulus intensity.

IV. SIMULATION AND STIMULATION DESIGN
According to the results of the first experiment, the area of tactile sensation on the arm was the smallest in the nerve bundle stimulation at the wrist. This indicates that tactile sensation induced at the unintended area by wrist stimulation is small. Furthermore, the electrode positions change the area of tactile sensation and the point of maximum sensation, particularly in the x direction on the hand in the first experiment. Therefore, nerve bundle stimulation on the wrist could potentially become a tactile display technology to induce sensation on the arbitrary position on the hand. However, the position of the tactile sensation should be controlled more finely to apply this method to induce tactile sensation in the hand.
To achieve this, we considered changing the current density distribution within each of the three nerve bundles (i.e., ulnar, median, and radial nerve), which innervate tactile sensation of the hand. Furthermore, to alter the current density distribution within each nerve bundle, the multi electrode wrist stimulation (MEWS) was proposed as Fig. 7d. In MEWS, 13 (E1 to El3) cathodal electrodes were attached around the wrist, and an anodal electrode was attached above the ulnar styloid process. Additionally, finite element analysis was conducted to demonstrate that MEWS changes the current density distribution within each nerve bundle.

A. NUMERICAL MODEL
The solid 3D male model (Zygote Media Group Inc.), which was constructed based on magnetic resonance imaging (MRI) and computed tomography (CT) and is consistent with anatomical knowledge, was used for the simulation. Furthermore, this model has also been used for educational materials. The model was imported into Scan IP (Simpleware, SYNOPSYS Inc.) which is the editorial system of the 3D model, and all parts of the model except for the left hand and arm were cut out, as shown in Fig. 7a. The elements of the model were classified into ten groups: bone, cartilage, vessel, skin, muscle, branches of median nerve, ulnar nerve, radial nerve, branch of radial nerve runs to front, and other nerves. There were unnatural gaps in the model considering the model did not contain blood and inner tissue. Therefore, these gaps were filled by the Boolean operation function of Scan IP. Thereafter, the filled region inside the vessels was classified as blood, while the remaining region was classified as inner tissue. Thirteen cathodal electrodes and an anodal electrode were manually attached to the skin model. Table 3 shows the center position of the surface of all electrodes.

B. SIMULATION OF CURRENT DENSITY DISTRIBUTION
The numerical model was exported as a NASTRAN format volumetric mesh file (2,407,065 elements in total), which was imported into COMSOL Multiphysics 5.6, a finite element analysis software, as shown in Fig. 7c. In COMSOL Multiphysics, the conductivity was assigned to each body part, as shown in Table 4 . These conductivities were defined based on the data base of IT'IS Foundation [30] and previous studies that simulated transcranial direct current stimulation [31].
The Laplace equation ∇ · (σ ∇V )(where V is the electrical potential and σ is the conductivity) was solved by applying the following boundary conditions: (i) Inward current = Jn (normal current density) was applied to the exposed surface of the anode attached on the wrist, (ii) the ground (cathode) was applied to the exposed surface of the electrode attached above the ulnar styloid process, (iii) all other external surfaces were treated as insulated, and (iv) the inward current density for each electrode was defined accordingly to adjust the current value to 3.0 mA, considering the superficial size of each electrode and observation from the first experiment that the mean value of the threshold current was up to 3.0 mA.
Thirteen stimulation conditions, namely, from 'E1' to 'E13' anodal stimulation, were simulated. Fig. 8a shows the current density distribution on the cross-section of each nerve bundle (i.e., ulnar, median, and radial nerve) when the wrist was cut in the plane formed by the centers of the E1, E5, and E9 surfaces. The red and blue colors show the high and low current densities, respectively. It should be noted that all parts that exhibited a current density higher than 0.12 A/m 2 are shown in red. Table 5 shows the coordinates of the maximum current density point on each nerve bundle and its current density. It should be noted that because Table 5 shows a cross-section of one of each nerve bundle, the maximum current density point may be located on a nerve bundle that does not have a higher current density in Fig. 8.

C. RESULTS AND DISCUSSION OF SIMULATION
The results of the simulation showed that the current density distribution formed on each nerve bundle relied on the electrode position. Furthermore, the areas of higher current density were located near the cathodal electrode position. For example, it could be observed from Fig. 8a that in   E1-E4 cathodal conditions, stronger current density distribution occured on the radial nerve, which was located near these electrodes. Similarly, under the E6-E7 cathodal and E8-E9 cathodal conditions, strong current density distribution occured on the median and ulnar nerves. Additionally, differences in the current density distribution among the conditions where the same nerve is stimulated, were observed. This means that the MEWS at the wrist can alter the current density distribution within the nerve bundle, which implies that the induced sensation could be modified within the innervating area of each nerve bundle.
Regarding the point of maximum current density in Table 5, some of the results were inconsistent with the above discussion. In particular, the maximum current density was generally low for the median nerve, and the maximum current density were the same value for median and ulnar nerves where median nerve was assumed to be stimulated the strongest. This itself implies that it would be difficult to only stimulate the median nerve. However, taking into account the results of the experiment from the previous section where the median and ulnar nerves were selectively stimulated with larger electrodes, median nerve is thought to be stimulated the strongest in these conditions. We considered that the current distribution on median nerves showed lower values because in the numerical model we used for simulation, the median nerve run closely to the skeleton where the conductivity is low and the current density distribution forms to avoid it.
Here, it was confirmed that the current density distribution formed on each nerve bundle relied on the electrode position, and the current density distribution within each nerve bundle was also modified.
Results of the first experiment and simulation showed that transcutaneous electrical stimulation induced tactile sensation generally on the hand and position of electrode had a major effect on the current density distribution on each nerve bundle. This indicated that multi electrode wrist simulation can potentially induce tactile sensation at an arbitrary position on only the hand. The second psychophysical experiment was conducted based on these findings.

V. SECOND EXPERIMENT: THE EFFECT OF MULTI ELECTRODE WRIST STIMULATION ON TACTILE SENSATION A. METHOD OF THE SECOND EXPERIMENT
This experiment was conducted to reveal the relationship between the position of the electrodes and the tactile sensa- tion position in MEWS to establish a method for controlling the position of tactile sensation on the hand.
Eleven healthy subjects (four females and seven males) participated in this experiment, which was conducted in accordance with the safety standards approved by the local ethics research committee at the University of Tokyo(21-117), Japan. The experiment was explained to the participants prior to their participation, and they were asked to sign a letter of consent. The study was performed in accordance with the ethical standards provided in the Declaration of Helsinki.
The experiment was conducted in a silent room. Ten cathodal gel electrodes (wizard gel, Yushiro Chemical Industry Co., Ltd) cut in (8 mm× 8 mm) were attached on the wrist at 12 mm intervals, whereas anodal gel square electrodes (Vitrode F-150S, Nihon Kohden Corporation) were attached above the ulnar styloid process. The starting position of this series of electrodes was at the back of the wrist between the index finger and thumb, which roughly corresponded to the position of E1 in Fig. 7, and each of the 10 electrodes was similar to the electrode position in Fig. 7. Electrode positions E1 to E10 were used based on the result of the simulation, which showed that there was no strong current density distribution area in the E11, E12, and E13 conditions. The subjects sat on a chair, and electrodes were attached on the left wrist. The distal end of the radius was used as an anatomical landmark for E1, and the electrode array was wrapped in the direction such that E10 was around the distal end of the ulna.
In this experiment, the same procedure was followed as in the first experiment for each electrode position. Fifty conditions of stimulation (10 electrode positions ×5 current strength conditions) were performed, and the subjects were asked to describe the tactile sensation induced by the electrical stimulation as conducted in the first experiment. The software used for recording the sensation differed from the one used in the first experiment in that only the wrist and hand were pre-drawn to observe the sensation in the hand in more detail (Fig. 9).

B. RESULT OF THE SECOND EXPERIMENT
In this experiment, one subject at E1 and six subjects at E10 did not perceive sensation at locations other than directly under the electrode. For this reason, these data were excluded for calculating the average and standard error of the maximum point of sensation.  Note that these averaged threshold values do not take into account the subject's experience with electrical stimulation, and the placebo effect which might have occurred during the measurement. Fig. 11 a shows the averaged area of tactile sensation for all cathodal electrode positions. It can be seen that a stronger current induces a wider area of tactile sensation. Additionally, the area of tactile sensation in E5 cathodal condition was the largest in this experiment. Error bars in this figure show the standard error. ART and ANOVA for the area of tactile sensation was conducted, and the results showed significant differences in the main effects of the electrode position factor (F(9, 90) = 15.9, p < 0.001, η 2 p = 0.24) and the current value (F(4, 40) = 53.70, p < 0.001, η 2 p = 0.32), whereas no significant difference was observed in the interaction effects (F(36, 360) = 1.15, p = .25, η 2 p = 0.084). Holm's multiple comparison tests for the position factor showed significant differences between E5 and all other positions, E3 -E9, E7 -E9, and E8 -E9. For the current factor, Holm's multiple comparison test showed significant differences between all combinations except for T × (1.4) 2 and T × (1.4) 3 . Fig. 11 b shows the averaged positions of the maximum tactile sensation position for all cathodal electrodes position conditions. The error bars show the standard error. The averaged coordinates of the point of maximum sensation for each stimulation position and intensity are shown as scatter graphs in Fig. 12. Each graph shows the x, y, and z coordinates of the maximum tactile sensation, which indicates that the position of the maximum tactile sensation shifted depending on the position of the cathodal electrode. Furthermore, ART ANOVA was conducted for the x, y, and z coordinates of the position of the maximum tactile sensation. While the results for the x coordinate showed significant differences in the electrode position factor (F(9, 90) = 215.85, p < 0.001, η 2 p = 0.81), no significant differences were found in the effect of the current value factor (F(4, 40) = 0.49, p = 0.74, η 2 p = 0.004) and the interaction effects (F(36, 360) = 0.42, p = 1.00, η 2 p = 0.032). Moreover, the Holm's multiple comparison tests for each position factor showed significant differences between all combinations except E1 -E4, E2 -E4, E6 -E10, E7 -E8, E7 -E9, E7 -E10, E8 -E9, and E9 -E10 (ps < 0.05). The results for the y coordinate showed significant differences in the electrode position factor (F(9, 90) = 8.59, p < 0.001, η 2 p = 0.15), whereas no significant differences were observed in the effect of the current value factor (F(4, 40) = 1.97, p = 0.10, η 2 p = 0.017) and the interaction effects (F(36, 360) = 0.40, p = 1.00, η 2 p = 0.031). The Holm's multiple comparison tests for each position factor showed significant differences between E5 and all other positions, E3 -E7 and E7 -E9 (ps < 0.05). Finally, the results for the z coordinate showed significant differences in the electrode position factor (F(9, 90) = 62.06, p < 0.001, η 2 p = 0.55), whereas no significant differences were observed in the effect of the current value factor (F(4, 40) = 0.66, p = 0.62, η 2 p = 0.006) and the interaction effects (F(36, 200) = 0.40, p = 1.00, η 2 p = 0.031). The Holm's multiple comparison tests for each position factor showed significant differences between all combinations except E1 -E9, E1 -E10, E2 -E8, E3 -E5, E4 -E6, E4 -E7, E6 -E7, and E9 -E10 (ps < 0.05).
To depict the relation of the electrode position and the position of maximum point of tactile sensation, Pearson's correlation coefficient between the x, y, and z directional positions of electrode in simulation (Sx, Sy, and Sz) and the x, z, and y directional averaged positions of the maximum tactile sensation (Ex, Ez, and Ey) was calculated. The results showed that there was high correlation between Sx-and-Ex, which was similar to thumb-little finger direction (r = 0.78). The correlation between Sy-and-Ey and Sz-and-Ez, which was similar to the finger-wrist direction and the palm to back of hand direction, was r = 0.20 and r = 0.38, respectively.

C. DISCUSSION OF THE SECOND EXPERIMENT
Results of this psychophysical experiment showed that our MEWS induced tactile sensation in the hand, and the position of the tactile sensation could be modified depending on the position of the cathodal electrode. However, one of the subjects did not perceive sensation at E1. Because E1 was placed at the distal end of the radius, which was the anatomical landmark, this subject could have a different location of the radial nerve from the majority of subjects around the wrist. Although individual differences in nerve location could be a limitation of this method, it could be minimized by properly calibrating the range of stimulation positions that induce sensation. Additionally, six subjects did not perceive sensation at E10. This is because of differences in the thickness of the wrists, which caused the E10 cathode to be located at where the nerve did not run. In this experiment, an electrode array with electrodes equally spaced was used for simplicity. An approach similar to the 10-20 system, which is a system adopted in the field of electroencephalography that uses several anatomical landmarks to divide the area of the cerebral cortex and standardize the placement of electrodes [32], could be considered to keep the consistency of electrode position and the underlying nerve among individuals.
According to the final questionnaire, most of the subjects described the perceived sensation as either vibration or paraesthesia-like sensation. This is consistent with the first experiment, which implies that stimulation position has small effect on what kind of sensation is induced. Fig. 10 showed that the tactile sensation was induced mostly in the hand and not in the arm in all cathodal electrode positions. Therefore, our MEWS can induce tactile sensation VOLUME 11, 2023 only in the hand area. These results were consistent with the results of the first experiment. Fig. 11 and the correlation between electrode position in the simulation and the maximum tactile sensation position indicate that the tactile sensation position on the hand moves depending on the electrode positions. These figures show that the tactile sensation moves from the thumb side of the back of the hand, through the palm, to the little finger side of the back of the hand.
According to Fig. 11 a, tactile sensation was reported in a wide area than the other position in the E5 stimulation. Together with the visualization in Fig. 10 and the y direction of the maximum point of sensation in Fig. 12 b, we can confirm that E5 stimulation induces tactile sensation over most of the area innervated by the median nerve, up to the fingertips. This infers that the stimulation at E5 activated more nerve fibers than the other conditions, and that this position was the closest to the course of the median nerve in this experiment.
On the other hand, the correlation between the position of the electrode and the tactile sensation in the palm to back of hand direction (z direction) was not high. There are two possible reasons for this result. One reason is that the way the height changes over the wrist and over the hand is different. In conditions where sensation occurred at the palm side surface of the hand, the z coordinate of the stimulation position changes because the wrist is modeled as an oval-like shape, while the z coordinate of the point of maximum sensation remains because the hand is a flat model. The second reason is that the z directional position of the tactile sensation elicited by E5 cathodal stimulation was lower than that of the adjacent stimulation E4 and E6 (Fig. 12 c). This is considered to be caused by the shape of the hand model we used for recording. The lateral view of the hand in Fig. 9 shows that the height of the hand surface decreases toward the fingertips, and that the z coordinate decreases as the y coordinate increases along the surface. Because E5 stimulation induced tactile sensation at the fingertips compared the adjacent positions, the z direction of the maximum point of sensation was exceptionally lower.
However, we can conclude from the results of this experiment that tactile sensations move from the palm to the back of the hand and from the thumb side to the little finger side, depending on the position of the cathodal electrodes. Overall, the difference of the current density distribution on the nerve bundles between the cathodal electrode positions actually occurs. This indicates that we can control the position of the tactile sensation more precisely by increasing the density of the electrodes and changing the distribution of the current density more precisely.

VI. CONCLUSION
This study demonstrated that the wrist is an effective position for stimulation in tactile sensation by TENS at the nerve bundles considering it eliminated the undesired tactile induction at the forearm and upper arm, which has been confirmed in several previous studies [12], [15]. Furthermore, we performed finite element analysis simulation based on the hypothesis that MEWS on the wrist can create a current density distribution bias within each nerve bundle. The second psychophysical experiment was conducted based on the simulation using the electrode arrangement that could actually create local current density distribution. The experimental results showed that our MEWS induced tactile sensation only in the hand and altered the position of tactile sensation by changing the position of cathodal electrode. Additionally, we confirmed that the positions of the maximum point of tactile sensation only existed on the hand. Moreover, the maximum tactile sensory position was not controllable in the distal direction, although it was controllable in the thumb-little finger direction. However, because presenting tactile sensation on the finger is one of the common applications for conventional tactile interfaces [25], this is a limitation in the proposed method.
The perceived sensation of our method was mostly described as vibration or paraesthesia-like sensation, and the  elicited sensation alone does not match natural tactile experience. Under this current limitation, we believe there are two main approaches to integrate this method with AR and MR technology. One approach would be to present tactile feedback depending on the activity in the AR/MR environment without interrupting the natural interaction. This is similar to a currently common VR technology which presents vibratory feedback to notify the user of activities such as touch. These feedback does not necessarily have to match the actual tactile experience. In AR or MR, because the users are assumed to be in contact with the real environment as well as the virtual environment, tactile feedback from virtual activities and interaction with the real environment must be compatible, which could be obtained with our method. It is known that changing the stimulation parameters can induce several kinds of distinguishable tactile patterns [14], which could potentially enable rich types of tactile feedback. The second approach is to modify the sensation by superimposing the sensation on natural tactile input. It is known that superimposing vibratory sensation on some interaction could modify VOLUME 11, 2023 the perceived sensation. By taking advantage of the fact that this method does not require covering the hands, modifying sensation could be one possible application to consider.
In view of the practical application of this method, future works in this field would include the following perspectives: 1. controlling the quality of sensation, and 2. dealing with the individual difference. First, in our study, we gathered information of what kind of sensation was induced only through questionnaires because our focus was only on the location of sensation. However, understanding and controlling the quality of sensation is equally important for application. Investigation on the relationship between the stimulation pattern and the quality of sensation would be the first step. Second, our MEWS was conducted by an equally spaced electrode array for simplicity. As a result, some of the subjects did not perceive the sensation at the edges of the array. Although it still successfully clarified the trends of our method, it is necessary to take into account the effects of individual differences such as wrist thickness for more precise control for practical application.
In conclusion, the novelty of our research is that we developed a multi electrode wrist stimulation method and demonstrated that tactile sensation can be elicited only on the hand without attaching electrodes and that the location of the tactile sensation can be altered from the palm to the back of the hand.  Tables 8 and 9.