An adaptive psychophysical procedure, Parameter Estimation by Sequential Testing (PEST) [28, 29], was used to determine perception threshold values for sensation in the hand or foot using a range of pulse duration and amplitude values. PEST is a method of threshold detection that was developed to obtain a more accurate estimation of sensory data and remove some of the variability and bias that occur when collecting data using subject reporting. The procedure consisted of starting with a subthreshold stimulus and increasing until a hand sensation was reported and then decreasing until the sensation disappeared. This was continued with a decreasing step size until the step size was below either 0.25V or 30µs depending on the parameter being varied. Based on the results of prior studies, efficient calculation of the strength duration curve can be performed using two trials of five threshold values [18]. PEST was used to determine the sensory threshold at two constant pulse durations and three constant pulse amplitudes using 0.5s pulse trains repeating every 1.5s. Two trials to determine threshold were performed in a random order under each of these five conditions and these sensory thresholds were used to calculate the strength-duration curve, the rheobase voltage and the chronaxie time as described by other [18, 30]. Briefly, the product of pulse amplitude (V_th) and duration (t) was plotted against pulse duration (t). This linear relationship was used to calculate rheobase amplitude and chronaxie time using Equation 1 Fig. (2a). The effective parameter space was defined with this strength-duration curve (Equation 2) as a lower boundary (blue line in Fig. (2b) and the maximum comfort level as an upper boundary (red lines in (Fig. 2b).

(1)

(2)

Fig. (2). Example strength-duration curve calculation from the ulnar nerve of subject 3. (A) The product of pulse amplitude (Vth) and duration (t) was plotted against pulse duration. This linear relationship was used to calculate rheobase amplitude and chronaxie time using Equation 1. (B) Resultant strength-duration curve obtained from Equation 2 (blue) with the same threshold points plotted as ‘x’. The upper red lines are the maximum comfort level from ramping at 100 µs (left portion) and 500 µs (right portion). The filled diamonds represent the 10 pulse amplitude and duration combinations used with the upper extremity nerves and the open squares are the 9 points used for the lower extremity nerves.

Ten locations within the effective parameter space (diamonds in Fig. (2b) were tested for the median and ulnar nerves with two pulse amplitude levels for each of five pulse durations (50, 100, 300, 500, 1000 µs). Nine locations were tested for the lower extremity (squares in Fig. 2b), with three amplitude levels for each of three pulse durations (50, 100, 500 µs). The trials were presented in the same random order for each subject and consisted of pulses at 50 Hz sent in repeating 0.5 second trains with 1 second in between. This frequency was chosen because higher frequencies were found to produce painful sensations [9]. For each set of stimulation parameters, subjects were instructed to report the modality, location, and intensity of the sensation in their hand or foot. A list of possible sensation modalities was provided (finger bent, pressing, weight bearing, probing, prickling, tingling, cool, numbness, buzzing, and unnatural) but subjects were also able to use a descriptive term of their choice. To standardize subject reporting, the palmar and dorsal surfaces of the hand were divided into 25 locations based on innervation patterns and expected sensation size Fig. (3). The dorsal and plantar surfaces of the foot were divided into 19 locations as seen in Fig. (3). Subjects were asked to point out the location of the sensation on a model hand or foot and the pre-defined location(s) that most accurately represented this area was selected in the computer by the experimenter. Intensity was recorded using an open-ended scale to allow relative comparison of strength. These values (modality, location and intensity) were recorded using a Matlab graphical user interface for later analysis.

Prior studies have reported an increase in the quality of the sensation through the use of non-constant or patterned waveforms [8, 25]. This is based on the fact that mechanoreceptor firing rate is rarely constant and varies across receptors [23]. In an effort to obtain more natural sensations, additional trials were performed using a number of patterned stimulation waveforms. The first two waveforms were based on promising results using electrodes implanted around the nerve [8]. All stimulus trains were sent at a pulse amplitude that was 60% of the range between the threshold at which the subject began to feel sensation in their distal limb and the maximum amplitude that the subject considered comfortable. This value was occasionally increased to obtain a more robust sensation using the lower frequency waveforms. Trials were randomized within each set of waveforms. Subjects were asked to state the sensation felt in their hand or foot after each stimulus train using the same sensation list as in the prior trials with the addition of ‘Tapping’. The modality, location and intensity were recorded on data sheets.

Full-scale and small-scale modulation were based on the equations presented by Tan et al. [8]. The pulse duration varied based on the base pulse duration (PD_base), the modulation frequency (f_mod) and the modulation percentage (M_per). The period of the pulses (time between the start of two subsequent pulses) remained constant at 0.02s. Equation 3 gives the relationship for the pulse duration, which is also shown in Fig. (4a). Actual waveform are shown in Figs. (4b and 4c). For full scale modulation, M_per was 1 (100%), PD_base was 500 µs, and f_mod varied between 1 and 2 Hz. For small-scale modulation, M_per was 0.2 (20%), PD_base was 500 µs, f_mod was 20 Hz. Tan et al used a much lower modulation percentage (about 5%) but this small fluctuation was not found to change the reported sensation during preliminary testing with surface stimulation so a higher percentage was chosen. Within these small scale modulation trials the burst length was varied from 0.05 to 0.5 s and the inter-burst time was varied from 0.05 to 0.9 s.

(3)

The final two waveforms involved a low frequency pulse train (1-4 Hz). In one of the waveforms, a pre-burst of 2-5 fast pulses (50 or 100 Hz) preceded the low frequency pulse train. This waveform was based on the firing pattern for slowly adapting mechanoreceptors which fire rapid pulses for a short time due to an indentation and then fire at a slower frequency while the indentation remains constant [23, 31]. The second waveform consisted of the low frequency pulse train (1-4 Hz) without the initial rapid pulses and was originally designed for the subjects to directly compare the two waveforms. The total waveform length was 5 seconds.

Fig. (3). Predefined sensation locations for the left hand and foot. There were 25 locations in the hand and 19 locations in the foot.

The purpose of the exploratory trials was to determine the effect of varying pulse amplitude and pulse width on modality, location, and intensity of evoked sensation. The percentage of times each sensation was reported for each nerve and each parameter was compared. Changes in intensity and location were assessed by comparing responses when the pulse amplitude was increased at a constant pulse duration.

The purpose of testing different waveforms was to find a waveform that produced realistic sensation with less paresthesia than found with the previously tested 50 Hz constant pulse train. For each nerve and waveform type, the occurrence of each sensation was quantified using the percentage of trials where the sensation was reported. Some similar sensations were grouped together, such as buzzing, tingling, prickling, and numbness (paresthesia). A McNemar’s Test was used to compare the different conditions and comparisons were considered significant if the family error rate was less than 0.05.

Fig. (4). The patterned waveforms utilizing non-symmetric, charge balanced pulse trains at 30 V (maximum recharge voltage of 4 V). (a) Description of how the pulse duration was modulated for a base duration of 1000 µs: Full scale modulation (blue) at a modulation frequency of 5 Hz and a modulation percentage of 100%; Small scale modulation (red) at a modulation frequency of 20 Hz and a modulation percentage of 20%. These are slightly different than the experimental values for demonstration purposes. (b) Resultant full-scale modulation waveform where the shortest pulse durations occur around 0.1 s. (c) Resultant small-scale modulation waveform has more subtle changes as the pulse duration modulates between 800 and 1200 µs.

The strength-duration curves used to define the lower bound of the effective parameter space are shown in Fig. (5). There was no significant difference between the rheobase (p=0.58) but there was a difference in the chronaxie time (p=0.03) due to the higher chronaxie in the lower extremity nerves. When compared to previously published data [18], there was no difference between the rheobase and chronaxie values for the median nerve (p=0.06) and ulnar nerve (p=0.12 for rheobase, p=0.45 for chronaxie). The lower extremity curves had more variability than the upper extremity and this was probably due to the fact that the protocol was changed half-way to force a threshold point at a higher pulse duration (500 µs). When the ANOVA was repeated using the final five subjects on each nerve there was no statistical difference (p=0.58). The first curves from the first five subjects are plotted in gray in Fig. (5c). The less accurate method could result in a lower rheobase which would lead to a higher chronaxie time since the chronaxie time is the duration threshold at twice the rheobase voltage. This possibly resulted in an underestimated threshold for the first 5 subjects, but since the minimum stimulation was at 25% of the range, foot sensation at this lowest level was reported in all but trial with one subject.

Fig. (5). Strength duration curves from each nerve with the average plotted in the thick red dotted line. (a) Median nerve; (b) Ulnar nerve; (c) Lower Extremity with first 5 subjects before the protocol change plotted in gray.

All subjects reported sensation in the foot or hand before muscle activation. This sensation was most often described as paresthesia (buzzing, tingling, prickling or numbness), and at least one of these descriptions was used by each subject. Just under half of subjects (30% from median nerve; 33% from ulnar nerve; and 64% from the lower extremity) reported more natural sensations, such as probing, tapping, or pressing at some point during the trials. In the subjects with an amputation, sensations such as pressure, toe bending, and pulsing were reported but paresthesia was once again reported most often.

At the lowest amplitude of stimulation for each pulse width (25% of the range), most upper-extremity subjects reported sensation in one or two of the pre-defined locations (median of 2; mean of 3; see blue bars in Fig. (6a). The number of locations ranged from 0 (no sensation felt) to 9, from UE subject 11 who reported sensation throughout the median nerve innervation area at the higher pulse durations. When stimulating at 75% of the range, the median number of locations increased to 4 with a mean of 3.9 (orange bars in Fig. (7a). Out of the 13 subjects who reported sensation in only one of the predefined locations, the most common locations were in the base of the palm (8 total subjects) and 5 subjects across both nerves reported sensation in only one finger tip. All reported locations from 25% and 75% of the range are shown in Figs. (7a and 7c) for median and ulnar nerve respectively. As the stimulus was increased, the number of reports of finger-tip-only sensation decreased as the number of reports of whole finger sensation increased. At 75% of the range, 80% of subjects reported sensation in the entire ulnar nerve innervation area while only 10% reported the entire median nerve area.

In the foot, just over half of subjects felt sensation in the toes at the lowest amplitude while the rest felt it along the lateral side of the foot. These two disparate locations are likely due to activation of the common peroneal or lateral sural cutaneous nerve in different subjects. The median number of predefined locations where sensation was reported was 1 for all three stimulation levels while the mean increased from 1.5 to 1.97 to 2.07 Fig. (7a). Since the predefined locations on the dorsal surface of the foot were large, an increase in sensation size may not have resulted in sensation in more than one predefined location. As the stimulation level increased, there were more reports of sensation on the top of the foot and along the lateral side (Fig. 7b).

As the pulse amplitude was increased, 76% of the trials on the median nerve had the same or greater area (expansion-Fig. (8a), 2% produced a smaller location (reduction) while the rest exhibited some loss in activation from the original area with the addition of new areas (relocation). Relocation varied from a slight loss of the original area to a complete change in the location of the sensation Fig. (8b). For the ulnar nerve, expansion occurred in 78% of trials and relocation was found in 16%. In the lower extremity, relocation was more common than in the upper extremity and found in 37% of trials while expansion occurred in 53%. This increase in relocation percentage for the lower extremity was likely due to the fact that the electrodes were near both the lateral sural cutaneous and common peroneal nerves and slight movements of the subject may have shifted the activation from one nerve to the other.

For the subject with an upper limb amputation, sensations were located near the center of the palm. Increasing stimulation did not change this location significantly. However, throughout the course of the stimulation, the subject reported warmth in the phantom limb, where previously the phantom had always been cold. This warmth persisted for several hours after testing concluded but no actual temperature of the limb was recorded. For the subject with a lower limb amputation, sensation started at the base of the big toe and expanded to involve all toes as the amplitude was increased. At higher levels of stimulation, a feeling of cramping in the toes was reported, similar to what was felt by this subject due to a touch on the top of the foot immediately prior to the amputation.

Fig. (6). Upper Extremity sensation location. (a) The number of predefined locations where sensation was felt during median and ulnar nerve stimulation. (b & c) The number of times each location was reported in the 50 trials for each stimulation level for the median (b) and ulnar (c) nerves. Multiple locations were reported in some trials so the total will be more than 50.

As the pulse amplitude was increased from 25% of the range of comfortable sensation to 75% of the range, the average normalized sensation magnitude increased significantly for both upper extremity nerves (p<<0.01). All three levels (25%, 50% and 75% of the range) were significantly different in the lower extremity (p<0.05, Tukey-Kramer method). When comparing individual subjects and increasing from 25% of the range to 75%, the magnitude was found to be higher or the same in 92% of pulse amplitude values from all three nerves. In the lower extremity, where there were three pulse amplitude levels at each pulse duration, the magnitude increased in 52 out of 60 comparisons between successive stimulation levels (87%). Similar results were seen in the subjects with amputations with all trial on the median nerve and half of the trial on the common peroneal nerve resulting in increased sensation magnitude. These results suggest that the magnitude of distal sensation can be at least crudely modulated using surface electrical stimulation.

Fig. (7). Lower Extremity sensation location. (a) The number of predefined locations where sensation was felt during lower extremity stimulation. (b) The number of times each location was reported in the 30 trials for each stimulation level. Multiple locations were reported in some trials so the total will be more than 30. The locations that arose from stimulating the lateral sural cutaneous nerve are depicted in purple.

Fig. (8). The effect of increasing pulse amplitude. (a) Example of expansion with activation of a greater area with increasing pulse amplitude, while maintaining activation in the original area. This result was from Subject 13 on the ulnar nerve stimulating at 300µs. (b) Example of relocation with activation of equal or greater area with differences in location. This result was from Subject LE10 stimulating at 100µs. The labels indicate the order in which locations were felt with increasing pulse amplitude (PA).