Abstract
Previous studies established that a salient collinear structure impairs local visual search. A display organization hypothesis states that the vertical grouping of elemental bars in the search display may selectively increase the salience of the local target in the background than that in the collinear distractor, leading to the collinear search impairment. Three displays were designed to test this hypothesis. A classical search display was adopted as a baseline. A diagonal search display was created with tilted bars, making perceptual organization diagonal and should reduce collinear search impairment. An illusory search display was designed by using abutting line illusion to emphasize the vertical grouping direction, which should increase collinear search impairment. A manipulation check was conducted with an online survey to understand the perceptual organization of the three displays. Results showed that the probability to perceive the stimuli grouping in the vertical direction was strongest in the illusory display and the least in the diagonal display. Nevertheless, the collinear search impairment did not vary with these manipulations, argue against the display organization hypothesis. We speculate that the search impairment might associate with the perceptual organization of the collinear distractor per se, rather than the perceptual organization of the background.
Visual search is usually enhanced by feature salience, which can be a unique feature in a display (Luck et al., 2021; Nakashima et al., 2015; Yantis & Jonides, 1984) or a well-grouped structure in a disorganized background (Baithwaite et al., 2005; Kim & Cave, 2001; Todd & Kramer, 1994), while learning effect are also involved (Chun & Jiang, 1998; Gaspelin & Luck, 2019; Shioiri et al., 2018; Tsuchiai et al., 2012). Such saliency could enhance target search if the target happens to overlap with a salient item, compare to the target non-overlapping with a salient item (Turatto & Galfano, 2001; Turatto et al., 2004; Yantis & Jonides, 1984). Nevertheless, the combination of orientation singleton and good continuity, which is “super salient” (Jingling & Zhaoping, 2008), counterintuitively impairs visual search (Jingling et al., 2017; Jingling & Tseng, 2013; Tseng et al., 2021; Tseng & Jingling, 2015). In particular, a local target that overlapped with a collinear structure is identified slower and less accurate than that non-overlapped with the collinear structure. This study aimed to test whether the search display organization contributed to the collinear search impairment.
The classical search display that obtained the collinear search impairment is shown in Figure 1. The display was filled by horizontal bars which arranged regularly. Meanwhile, one of the columns was filled with vertical bars and this column is the collinear distractor. This collinear distractor had bars with different orientation to the other bars (orientation singleton), while the vertical bars aligned head-to-tail to each other formed a collinear grouping. Therefore, the collinear distractor is salient (Jingling & Zhaoping, 2008). A target (Figure 1C) was placed on one of the bars, which can be in the collinear column (overlapping, Figure 1A) or in the background (non-overlapping, Figure 1B). The probability of overlapping condition was designed to be at a chance level so that the collinear distractor is task-irrelevant. Since the collinear distractor is salient, and the target was occasionally overlapped with the collinear distractor, then target discrimination should be easier or faster for targets overlapping with the collinear distractor (Figure 1A) than the non-overlapping cases (Figure 1B). Nevertheless, search impairment was observed by worse performance for the overlapping than the non-overlapping conditions, contradictory to the findings in the classical attentional capture literatures (e.g., Turatto & Galfano, 2001; Turatto et al., 2004; Yantis & Jonides, 1984).
The examples of search display used in Jingling and Tseng (2013) . Note: (A) Part of the search display of overlapping condition. (B) Part of the search display with a non-overlapped target. (C) Examples of the possible targets.
The collinear search impairment is a robust phenomenon. First, the search impairment did not specifically associate with orientation of the bars of the target. For instance, Jingling and Tseng (2013) have checked the condition in which displays in Figure 1A and B are rotated 90°, such that the non-targets are vertical bars and the collinear distractor a horizontal raw. Still the collinear search impairment was found. Also, Liu and Jingling (2021) have checked the orientation discrimination performance on a target (Figure 1C, the tilted break) on either a vertical or a horizontal bar. Results showed no significant differences between orientation of the target bars. Further, the search impairment was associated with the grouping strength of the collinear distractor. The longer (more bars aligned to each other in the collinear column) the collinear distractor was, the larger the search impairment can be observed (Chow et al., 2013; Jingling & Tseng, 2013). In addition, the search impairment was uniquely associated with collinear grouping. Jingling et al. (2013a) showed that a distractor column with a unique color or luminance did not lead to search impairment. Curving the collinear distractor so that bars on the distractor were not the same in orientation can still lead to search impairment (Lu & Jingling, 2017). Together, the characteristics of the collinear distractor associated with the collinear search impairment.
One possibility of why the collinear search impairment occurred might be related to the perceptual grouping of the ends of the bars in the background. Figure 2 shows how the perceptual grouping of ends of the bars may lead to search impairment, which is called the display organization hypothesis hereafter. In the classical display, the horizontal bars arranged regularly in the display, so as the ends of the bars. If the display contained only horizontal bars, the display organization should be horizontal. However, in the classical display, since there was a collinear distractor grouped within a column in the display, the ends the horizontal bars may be induced to group in the vertical direction (Vickery, 2008). Participants may perceive the ends a vertical grouping (indicated by the red rectangular in Figure 2), which can be considered a weak form of the abutting line illusion (Soriano et al., 1996) that was induced by aligned ends of the stimuli, though the abutting line illusion was usually produced by interleaved ends of bars. In this case, the horizontal bars would be grouped into ladder-shape columns, while a target in such a ladder-shape column would become a gap singleton (Figure 2B). Together with the homogeneous of the environment (Duncan & Humphreys, 1989; Feldmann-Wüstefeld & Schubö, 2015), the non-overlapping targets should become salient. Meanwhile, in the overlap condition (Figure 2A), the target becomes one of the gaps between bars since the target and the gaps were similar in sizes. In other words, non-overlapping targets may be more salient than the overlapping targets, leading to a faster and more accurate search for non-overlapping than overlapping targets (i.e., the collinear search impairment). The display organization hypothesis may also explain the observations in Jingling et al. (2013a), in which the proportions of short latency saccades (could be taken as an index of perceptual salience, van Zoest et al., 2004) were less for the overlapping targets than non-overlapping targets.
The Illustration of the display organization hypothesis. Note: (A) and (B) are the classical display, while (C) and (D) are the diagonal display. The red boxes highlight the possible grouping directions.
To investigate whether the display organization hypothesis could explain the collinear search impairment, two types of displays were designed: the diagonal display and the illusory display (Figure 3). The purpose is to reduce the strength of the vertical or columnar grouping in the former and to enhance the vertical grouping in the latter, respectively. If the search impairment associated with this subjective vertical grouping, then the collinear masking effect should be smaller in the diagonal display and be larger in the illusory display, respectively, compared to the baseline of that in the classical display. If, however, a local target was still more difficult to discriminate when it was overlapped with the collinear distractor compared to non-overlapped, regardless of the types of search displays, then the collinear search impairment could not be fully attributed to the grouping based on display organization.
Examples of the search displays used in this study. Note: (A) A right-tilted target is overlapping with the collinear distractor in the classical display. (B) A horizontal target is overlapping with the collinear distractor in the diagonal display. (C) A left-tilted target is non-overlapping with the collinear distractor in the illusory display.
First, the diagonal display (Figures 2C, 2D, and Figure 3B) was designed to reverse the relative salience between the overlapping and the non-overlapping targets from the display organization. In the diagonal display, all the bars were tilted in one direction (e.g., 135° in Figure 2C) while one group of bars, crossing different columns, tilted in another direction (e.g., 45° in Figure 2C), creating a salient collinear distractor in the diagonal direction. Since bars were still regularly aligned in invisible grids, the ends of the bars could be grouped in the vertical or horizontal directions. Nevertheless, the dominated perceptual grouping should be in the diagonal direction, following the salient collinear distractor. If the display was organized in the diagonal direction, then, as shown in Figure 2C, the overlapping target was still the gap singleton, while the non-overlapping target would not. In this case, the overlapping target should be more salient than the non-overlapping targets. Therefore, the display organization hypothesis predicts that overlapping targets should enjoy attention priority more than non-overlapping targets, leading to better performance for the former than the latter in the diagonal display.
On the contrary, the illusory display (Figure 3C) was designed to exaggerate the perceptual grouping in the vertical direction. Taking advantage of the abutting line illusion, the vertical position of the columns with horizontal bars alternated their height levels to create a stronger abutting illusion in the vertical direction. According to Soriano et al. (1996), the bars used in the search display are in the optimal width range (> 0.2° visual angle), number range (more than 7 bars), and orientation range (zero degrees alignment) to induce this illusion. Although the length of the bars is rather short (our bars were 0.91° in visual angle, while this illusion would be stronger with bars in more than 4° visual angle long), our texture-like search display may enhance such repeated pattern and create a strong impression of vertical grouping. In this case, the tendency to group horizontal bars into columns is expected to be higher in the illusory display (Figure 3C) than in the classical display (Figure 3A), which perhaps enhances the salience of the non-overlapping targets. The prediction is that, if the search impairment was induced by the perceptual grouping of the display organization, then the search impairment should be larger in the illusory display than that was observed in the classical display.
Method
The study invited participants to complete three search tasks using the three designs of the displays (Figure 3): the classical display, the diagonal display, and the illusory display. Since the target in each display can be different, the three displays were shown in different blocks. Additionally, a survey was designed to verify the assumed perceptual grouping for the three search displays.
Participants
According to our previous study, the search impairment is a robust phenomenon and can be observed with 12 participants (Jingling & Tseng, 2013). We used G*Power 3.0.8 (http://www.gpower.hhu.de/) to estimate sample size. The partial eta square was 0.28 in Liu and Jingling (2021), by using F tests ANOVA repeated measures within factors, one group sample size of 12 can reach 0.05 alpha level and 0.95 power level. Finally, we collected 15 participants. There are 11 females and 4 males, and their average age is 23.37 years old and the range was from 20 to 29 years old. They all self-reported to have normal or corrected-to-normal vision and signed the inform consent approved by China Medical University & Hospital Research Ethics Center.
Equipment and Stimuli
The experiment was carried out in a desktop controlled by E-Prime 2.0 professional in a dimly room. Stimuli showed on a 24′′ LCD flat screen with resolution 1024 × 768. Participants viewed the screen from 70 cm away with a chin-rest.
Figure 3 shows the examples of the search display used in this study. The search display was composite with 9 bars by 13 bars, each extended 0.91° by 0.15° in visual angle. Bars were placed at the center of grids in 1.05° visual angle, making the whole search display about 9.45° by 13.65° in visual angle. The target was 0.3° by 0.15° in visual angle, presented on one of the possible locations at the 3rd (e.g., Figure 3A), 5th (e.g., Figure 3C), 9th (e.g., Figure 3B), or 11th columns. The vertical location of the target was fixed at the center row (the 5th row). In the classical and illusory search display, the target was a tilted black bar as that was used in the classical display (Figure 1C), either left-tilted (e.g., Figure 3A) or right-tilted (e.g., Figure 3C) in 45°. In the diagonal display, the target was a horizontal (e.g., Figure 3B) or vertical black bar. Between trials there was a fixation display, which was a blank screen with a white dot in 0.5° visual angle at the center.
Design
The experiment is a two-factor within design, while the factors are the display type (classical, diagonal, and illusory display) and the target type (overlapping or non-overlapping with the collinear distractor). The classical display (Figure 3A) was similar to that was used in our previous studies (e.g., Jingling & Tseng, 2013), excepting that the display contained only 9 by 13 bars. The diagonal display (Figure 3B) contained bars oriented either 45° or 135°. Among them, bars on one diagonal were in the other orientation (e.g., 45° in Figure 3B) compared to the rest of bars in the search display (e.g., 135° in Figure 3B), forming a salient collinear structure, which is the collinear distractor. Due to the size of the display, those collinear distractors passed the 3rd and 11th columns would be 7 bars long, while those passed the 5th and 9th columns would be 9 bars long (e.g., Figure 3B). The illusory display (Figure 3C) alternated horizontal level of each column in a size of half grid (∼0.53° in visual angle) in the classical display, making the boundary between columns formed the abutting line illusion. The two columns next to the collinear distractor can be either the same height (e.g., Figure 3C) or different height. The collinear distractor did not change its height level. The target was overlapping with the distractor in one-fourth of the trials, making the distractor and the target task-irrelevant.
The display type is block designed, that is, each search display was carried out separately in a section. The target type is a completed randomized design. Each section contained 128 trials, which is the combination of two orientations of the target, two orientations of the background bars, four possible target locations, four possible distractor locations, and two repetition. These trials were presented in two blocks, each is one repetition, and all the conditions were completely randomized in their presentation sequence. The sequence of sections was counterbalanced between participants.
Procedure
Each trial started with the fixation display. After 800 ms, the search display was shown until response. After that the next trial started immediately. In the classical and illusory display sections, participants pressed the “f” key for a left-tilted target and the “j” for the right-tilted target. In the diagonal display section, participants pressed the “0” key for a vertical target, and the “2” key for a horizontal target. Both speed and accuracy were emphasized. Twelve trials were randomly chosen for practice before each section. The whole procedure took about 40 min.
The survey
A manipulation check was accessed by an online Google form. The survey was collected in two undergraduate courses of Introduction to Psychology as part of demos of perceptual grouping when lecturing visual perception. The students did not get any course credits by doing this survey, and the survey did not record any participant information. The students of these two courses were not overlapped since the same courses cannot be repeatedly registered. There were ten questions in the survey. Participants were asked in which direction if you want to cut the figures with a scissors according to the layout of the features: vertically, horizontally, or not above directions. They were encouraged to follow their intuition. Two figures were formed by round discs which should be grouped vertically or horizontally. The rest figures were as follows: homogeneously vertical bars, homogeneously horizontal bars, homogeneously diagonal bars, homogenously illusory background bars, the classical display, the diagonal display (6 by 6 bars, the small version), the diagonal display (9 by 9 bars, the large version), and the illusory display. The classical and illusory display used in the survey were all large version (9 by 9 bars). The sequence of these 10 figures were presented randomly for participants. It took about 2–5 min to complete the survey.
Results
Response time, accuracy, and BIS
The response times (RT) that were faster than 200 ms (anticipation responses) and slower than one's own three times of standard deviation (too slow responses) were excluded for further analysis. As a result, 1.71%, 3.70%, and 3.49% of trials were removed for the classical, diagonal, and illusory displays, respectively. To consider response time and accuracy together, we calculated the balanced integration score (BIS, Liesefeld & Janczyk, 2019) by equally weighting the normalized RTs and accuracy. The BIS is less sensitive to the speed-accuracy trade-offs of the participants may adopted compared to the traditional methods. To calculate BIS, the RTs of all participants in every condition were listed and transferred to normalized z score, so as the accuracy. Then the BIS equals to the differences between the normalized RT and the normalized accuracy. A positive BIS means that the condition is relatively easier (i.e., participants can respond faster and more accurately), and a negative number indicates that the condition is more difficult among all the conditions. Table 1 shows the results of the three tasks.
The descriptive statistics of the response time (ms), accuracy (%), and the balanced integration score (BIS).
Note: RT is response time, BIS is the balanced integration score (Liesefeld & Janczyk, 2019), SE is the standard error of the mean.
A two-way repeated-measure ANOVA was carried out for the three dependent variables, respectively. Results of reaction times showed a significant main effect of search display, F (2, 28) = 28.62, ηp2 = .67, p <.001. Tukey post-hoc analysis showed that the response was longer for the diagonal display (875 ms) than the illusory display (643 ms), and the RT was longer for the diagonal display than that for the classical display (625 ms), ps <.01. Also, the main effect of target type was found, F (1, 14) = 56.89, ηp2 = 0.80, p <.001. That is, responses were longer for overlapping targets (745 ms) than non-overlapping targets (683 ms). Interestingly, there was no significant interaction, F (1, 14) = 1.32, p = .29. In other words, the search impairment induced by the collinear distractor was observed in all three types of displays, while the sizes of the effect between displays did not reach statistical significance.
Accuracy (Table 1) was also submitted to ANOVA and found a main effect of target type, F(1, 14) = 6.25, ηp2 = 0.30, p = .03. Thus, participants made more errors for overlapping targets (92.81%) than non-overlapping targets (95.50%) regardless of types of the displays. The main effect of search display (p = .20) and the interaction between display and target types (p = .91) were not statistically significant. Thus, search impairment induced by the collinear distractor was again obtained in accuracy in general regardless of types of search displays.
Finally, the results of BIS were also submitted to ANOVA. Table 1 shows the BIS in each condition. A significant main effect of display type was found, F (2, 28) = 14.45, ηp2 = 0.51, p <.001. Post-hoc Tukey test showed that BIS was the smallest for the diagonal display (−1.15) than that for the illusory (0.34) or classical (0.82) displays, ps <.01, respectively, suggesting that the diagonal display condition was more difficult compared to the other two conditions. Also, the main effect of target type was found, F (1, 14) = 19.59, ηp2 = 0.58, p <.001. The BIS for overlapping condition (−0.40) was smaller than that for non-overlapping condition (0.40), confirming the collinear search impairment. The interaction between search display and target type again did not reach statistical significance, p = .81. The BIS data was plotted in Figure 4 by separately presenting the BIS in the overlapping and non-overlapping conditions for the three displays. The results showed that the collinear search impairment was significant in all three displays and the size of the collinear search impairment did not vary significantly across search displays.
The balanced integration score (BIS) for the overlapping and non-overlapping conditions in the three search displays. Note. The error bars are the standard error of the mean.
Results of the survey
Finally, 82 undergraduates completed the survey. The results were shown in Table 2 and some selected results were shown in Figure 5. The display with horizontal bars was mainly perceived as grouped in the horizontal direction (87.80%), while the display with vertical bars was mainly perceived to be grouped in the vertical direction (91.46%), validating the survey. The tendency to perceive the display organized in the vertical direction was higher in the classical display (36.59%) than that in the homogeneously horizontal-bar display (10.98%), supporting the idea that the collinear distractor may induce a tendency to group items in the vertical direction (Vickery, 2008),χ2 (2) = 25.02, p <.001. Most interestingly, the proportion of perceiving vertical organization varied with the types of the search displays: 36.59% for the classical display, 12.20% for the diagonal display, and 59.76% for the illusory display (Figure 5). The differences between the three conditions were statistically significant by Chi-square independent analysis,χ2 (4) = 107.55, p <.001. Therefore, the three displays used in current study generated different subjective grouping: the diagonal display generated the least tendency in vertical grouping, while the illusory display elicited the strongest tendency in vertical grouping. This ranking matched with what was assumed in the display organization hypothesis.
The rating proportion of the perceived grouping for different displays. (A) The display with horizontal bars; (B) the display with vertical bars; (C) the classical display; (D) the diagonal display; and (E) the illusory display.
The frequency (number of participants) of ratings for ten testing figures in the manipulation check survey.
Discussion
This study aimed to examine whether the display organization hypothesis can explain the collinear search impairment. The display organization hypothesis assumed that participants may perceive the classical display in a columnar manner (gaps between bars were grouped vertically, Figure 2A), and thus alter the relative salience between the overlapping and non-overlapping targets. Following the hypothesis, the collinear search impairment should be stronger in the illusory display than that in the classical display, and should be weaker in the diagonal display than that in the classical display. The data of this study confirmed that the tendency to perceive vertical grouping varied between displays (Figure 5); nevertheless, the size of the collinear search impairment was not significantly different to each other in the three types of displays (Figure 4). Our data therefore argue against the display organization hypothesis for the collinear search impairment.
Our manipulation check (Figure 5) confirmed that the display organization varied across the three displays; however, these grouping differences between displays did not reveal in target discrimination (Figure 4). As shown in Figure 5, interleaving the horizontal bars in the illusory display, compared to that in the classical display, significantly increases the percentage of perceiving vertical organization from 36.59% to 59.76%. Meanwhile, the diagonal display created less tendency for vertical grouping (12.20%). The display organization hypothesis predicted that the collinear search impairment should vanish in the diagonal display because the overlapping targets should be more salient than the non-overlapping targets. Nevertheless, our data showed that reliable collinear search impairment was still observed in the diagonal display. Also, the display organization hypothesis predicted that the collinear search impairment should be larger in the illusory display than that in the classical display. There was no evidence supporting for the prediction. Therefore, the display organization hypothesis was supported in that the manipulations on the three displays did vary the tendency of vertical grouping, but was not supported in the experimental data in that the collinear search impairment did not vary according to the percentage of vertical grouping. In this case, we concluded that the collinear search impairment cannot be explained by the display organization hypothesis.
Our findings suggest that the collinear search impairment is not determined by display organization. The current finding seems to be contradictory to the observations in which decreasing perceptual grouping strength of the search display can reduce the collinear masking effect (Liu & Jingling, 2021). However, it is notable that in Liu and Jingling (2021), not only bars in the background but also bars in the collinear distractor increased separation between bars. Therefore, we speculate that the collinear search impairment might associate with the perceptual grouping of the collinear distractor per se, rather than the perceptual grouping of the display. As reported in previous studies, the search impairment increased with number of bars in the collinear column (Jingling & Tseng, 2013), and disappeared when the distractor column was not defined by collinearity (Jingling & Tseng, 2013; Jingling et al., 2013b; Lu & Jingling, 2017). Our findings implied that the collinear search impairment might have been observed in an irregularly distributed display as long as the distractor was collinear. As Lu and Jingling (2017) showed, the collinear search impairment was observed in the display with randomly oriented bars. Future study can test whether bars in the background were with randomized locations, or even without bars in the background, can also produce the collinear search impairment.
Our findings also provide one more piece of evidence on that the collinear search impairment is not associated with perceptual salience. The display organization might help to increase perceptual salience of the non-overlapping targets than that of the overlapping targets in the classical display (Figure 2; Duncan & Humphreys, 1989; Feldmann-Wüstefeld & Schubö, 2015; Jingling et al., 2013a); however, reverse the relative salient in the diagonal display did not therefore eliminate the search impairment. This observation is consistent with the findings in Jingling et al. (2017), in which making the target per se more salient did not eliminate the collinear search impairment. Also, one may consider that the alternation of the horizontal height level of the background bars in the illusory display not only alters the display organization but also contributes to the salience of the collinear distractor, leading to a more salient collinear distractor in the illusory display compared to that in the classical display. However, collinear search impairment was observed when the collinear distractor was less salient (Lu & Jingling, 2017; Jingling et al., 2017) than that in the classical display, once again implying that the collinear search impairment is independent to perceptual salience.
Our data also supported that the collinear grouping per se is not enough to induce collinear search impairment, rather, the collinear distractor needs to combine collinear grouping within the distractor and orientation contrast to its neighbors to elicit the masking effect (Jingling et al., 2013b). For instance, the horizontal bars in the background in the illusory display were not collinearly grouped as that in the classical display, while this change did not alter the size of the collinear search impairment. Therefore, whether the bars in the background were collinearly alignment is not crucial, probably because neighboring bars were all horizontal and no orientation contrast in the area. These parallel horizontal rows in the classical display played similar roles as the interleaved horizontal rows in the illusory display.
One of the limitations of the study is that the targets were not the same in the three displays. Due to the bars were all slant in the diagonal display, the target changed to be a vertical or a horizontal gap on a bar, rather than a left- or right-tilt gap as in the classical display. The BIS results (Figure 4) also confirmed that the diagonal display was the most difficult condition among the three. Nevertheless, regardless of the difficulty, differences between overlapping and the non-overlapping conditions remained the same. Another challenge is that we did not have an independent manner to evaluate the salience of the small target in a display with a more salient global structure (the collinear distractor). Accuracy in each condition can be considered an approximate estimation of the salience of the target (e.g., Nothdurft, 2000). Since the accuracy of the overlapping targets (92.81%) was lower than non-overlapping targets (95.50%) regardless of types of the displays, we considered that the overlapping targets were less salient than non-overlapping targets. Perhaps perceptual salience in such complex display took different weightings from difference feature contrasts, or the collinear grouping also contributed to salience perception (Jingling & Zhaoping, 2008).
As to why a collinear grouped structure masked a local target is still unknown. One possibility is that the target as a gap on one of the bars, which is too similar to these gaps between bars, thus lead to mis-identification of the target. Nevertheless, a shape target (Tseng & Jingling, 2015) or a brighter/dimmer target (Jingling et al., 2017) still being masked by the collinear distractors. Alternatively, the collinear grouping via lateral connections may induce perceptual filling-in (Zhaoping & Jingling, 2008) and smear the target visibility; however, Tseng et al. (2021) found that lowered down the luminance of bars to the threshold level (which should avoid collinear grouping via lateral connections) did not remove collinear search impairment. Another possibility is that the collinear grouping may provide location information that interferes location uncertainty of attentional distribution (Shioiri et al., 2016), which might lead to reduce uncertainty to the non-overlapping target more than overlapping targets. Further studies are needed to explore these possibilities.
In conclusion, the current study showed that the display organization did not cause the collinear search impairment in visual search. Further study should focus on how perceptual grouping of the collinear distractor could mask a local target. Our study may provide some hints for camouflage.
Footnotes
Acknowledgements
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Nationwide Cooperative Research Project, RIEC, Tohoku University, Ministry of Science Technology Taiwan, Ministry of Science and Technology Taiwan, (grant number H29/A24, MOST106-2420-H-039-002-MY3, MOST110-2410-H-039-005).