Each subject was able to perform the desired movement by using the index finger with reasonable accuracy. The data of all subjects were therefore included in the following analysis.
4.1 Results of behavioral test
All subjects reported that the tactile stimulus deduced by passive mode was more intense than those by active mode, and the rated similarity of active sliding was significantly greater than passive sliding (t(13)=7.16, p<0.001, Mactive=7.64, SD=0.74; Mpassive=4.14, SD=1.61). Table 1 showed the rating scores for the similarity to the daily experience, produced by the active and passive sliding.
Table 1 Result of questionnaire survey
Participants
|
Which sliding mode gives the stronger feeling
|
Rating score for active sliding
|
Rating score for passive sliding
|
1
|
Passive sliding
|
8
|
3
|
2
|
9
|
2
|
3
|
8
|
6
|
4
|
7
|
5
|
5
|
8
|
3
|
6
|
7
|
5
|
7
|
8
|
7
|
8
|
8
|
2
|
9
|
8
|
5
|
10
|
6
|
3
|
11
|
7
|
5
|
12
|
8
|
6
|
13
|
8
|
3
|
14
|
7
|
3
|
4.2 Results of the friction test
Fig. 5 compares the fNIRS and friction measurements obtained from a single participant sliding against PTFE#40 for one trial. The changes of hemoglobin amplitude measured in CH14 are displayed in the graph, showing the relatively stable HbR signal and the fluctuant HbO signal. The friction measurement including the normal force and friction force were presented, comprising three periods of reciprocating motion. Participants were asked to keep the normal force at 1.5 N. It could be observed that the normal force measured in passive sliding (Fig. 5(b)) was more stable than in active sliding ((Fig. 5(a)). Compared to the irregular amplitude variation of hemoglobin signal during the task, the friction force was rhythmical with the change of sliding direction.
Fig. 6 displays a typical friction measurement for one period of reciprocation captured in Fig. 5 (green dashed area). During the test, participants slid their finger first in the proximal direction, then in the distal direction, performing the reciprocating motion, the normal force was controlled by participants through observing the real time signals. The graph shows the measured friction force, normal force and friction coefficient. It could be observed that the normal force was fairly constant at 1.5 N, the values of friction force and the friction coefficients were positive in the proximal direction and negative in the distal direction. In active sliding (Fig. 6 (a)), participants would change the direction and move in the opposite direction immediately upon finishing in the proximal direction. In passive sliding (Fig. 6 (b)), the motion of the sample table was controlled by the servo motor and there was a brief stationary state for the device to change the moving direction. The transition states were also shown in the contact force and friction coefficient signals.
The distribution of friction coefficients from 14 participants measured under four experimental conditions are shown in the boxplots of Fig. 7. The average value of friction coefficients was calculated from the stable sliding region. The square marker indicates the average value, while the horizontal line indicates the sample median. The size of the box corresponds to the first and third quartiles. The error bar indicates the total range of values obtained, excluding any outliers. Outliers are marked as small black diamonds. In both two sliding modes, the distribution of friction coefficients measured on the rough sample was lower than that measured on the smooth one. For the same surface roughness sample, the distributions of friction coefficients were similar in active and passive sliding. By analyzing the effect of surface roughness, the paired sample t-test showed a significant difference both in active sliding (t(13)=-6.21, p<0.01) and passive sliding (t(13)=-2.83, p<0.05), indicating that the friction coefficient decreased with the increase of surface roughness. By analyzing the effect of sliding mode, there was no significant difference between the rough and smooth samples (p>0.05).
4.3 Results of fNIRS test
Fig. 8. shows the average measurement of HbO and HbR collected in CH7 for six trials from a participant moving against PTFE#40 with the two sliding modes. The sampling data were presented from -2 to 50s, the sliding task duration was 0-20s as indicated by the dotted lines. It can be observed that during the task, the amplitude of HbR remained relatively stable while that of HbO fluctuated significantly. The amplitude of HbO increased obviously at the beginning of the sliding, then trended to decrease at the end of the task, meaning that HbO can reflect effectively the cerebral blood flow during the neural activity. In addition, the amplitude of HbO measured in passive sliding (Fig. 8(b)) was higher than in active sliding (Fig. 8(a)), indicating a higher level of activation occurring in CH7 with passive movement.
The HbO data from 14 participants were superimposed and analyzed statistically. Two regions of interest (ROIs) were analyzed by using 2×2 (sliding modes × surface roughness) repeated measures ANOVA. The first ROI was the pre-motor and supplementary motor cortex, including three channels, such as CH2, CH8 and CH10 as shown in TableR1 in Supplementary file. The ANOVA results shows that there was neither a significant main effect of the sliding modes, surface roughness nor were their relevant interactions (p>0.05), indicating that there was no activation in the pre-motor and supplementary motor cortex. The second ROI was the dorsolateral prefrontal cortex, including CH7, CH14 and CH18. The ANOVA results revealed a main effect of the sliding modes (F(1,13)=9.190, p<0.05, η2p=0.231), and neither a significant main effect of the surface roughness nor were their relevant interactions (p>0.05).
To further investigate the effect of sliding modes in DLPFC, the paired sample t-test was employed between finger active and passive sliding, revealing a main effect of sliding modes occurred in CH7(t(27)=4.392, p<0.01), CH9(t(27)=4.172, p<0.01), CH12(t(27)=4.141, p<0.01) and CH14 (t(27)=2.964, p<0.05) while no significant difference in other channels (p>0.05). The positive t value measured in activated channels indicates that a passive sliding stimulus produced more positive activation effects in relevant regions than an active stimulus, meaning a higher activation level for passive sliding mode, corresponding to the assumption. Fig. 9 displays the activated channels between finger active and passive sliding modes on a head model, a highly significant difference occurred in CH7, CH9 and CH12. Table R1 in the supplementary file shows that CH7 includes the dorsolateral prefrontal cortex and frontal eye fields; CH9 is the frontal eye fields; CH12 is the frontopolar area; CH14 includes the dorsolateral prefrontal cortex, frontal eye fields and frontopolar area. Accordingly, three cortex regions, dorsolateral prefrontal cortex, frontopolar area and frontal eye fields were mainly activated with the two sliding modes.