Efficiency and strategy of visual search: a study on eye movement and scan path

doi:10.21203/rs.3.rs-1732929/v1

Download PDF

Article

Efficiency and strategy of visual search: a study on eye movement and scan path

https://doi.org/10.21203/rs.3.rs-1732929/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Search efficiency is related to page design, which can be reflected by search time, and is related to eye movement. Eye-tracking metrics can not only affect the search efficiency, but also reflect the search strategy. In scan path, the line segments between fixations form scanpath angles. Each angle represents the direction of eye movement and the strategy to determine the location of the next fixation. We created SAM (scanpath angle model), divided all angles into four intervals, and calculated their percentages to better understand these angles. Two eye movement experiments were designed. The first is to search for targets in non-distracted and distracted pages, and the second is to search for targets in pages with different layouts. It was found that distraction decreased search efficiency while increasing fixation count, fixation duration, and saccade count. The above three metrics have no significant different when layout changed, but the layout of scan path matched the page layout. The scanpath angle between 0 and 45 degrees is the largest in the distraction and layout groups, and the percentage is quite close.

Visual search refers to an individual finding a specific target in a specific area, which generally involves dynamic eye movement.^1,2 The user centered design must consider user's searching behaviors to improve user satisfaction³. Search efficiency is critical for users. The process of visual search includes both top-down and bottom-up, and the form is very flexible⁴. Many factors affect users' search behavior and search efficiency. The size of graphics and text affects the search efficiency⁵. The higher the difference between target and non-target, the more efficient the search; the lower the difference between non-targets, the more efficient the search⁶. Increasing the number of colors will reduce the efficiency of searching icons, while increasing the variations between elements improves it.^7,8. Users’ search efficiency is also related to the visual complexity of the interface^9–12. The increase of the number of items (including target and interference) in the interface will lead to the increase of search time^13–15. And the layout of the interface can impact search efficiency and user satisfaction of the website^16,17. Search efficiency is not only affected by the structure of the interface and the environment, but also related to the user's search strategy^18–20. The scan path represents the eye movement sequence, which is related to the type of search task and users' search habits, and affects the search efficiency^17,21,22.

Eye tracking is an objective method of measuring visual attention that can be used to study user's search efficiency and strategy^{11,17,18,22−24}. The results of eye tracking studies can also help designers improve product interactivity²⁴. The behavior of individuals observing and searching for targets in the graphical interface includes fixation and saccade, which appear alternately. Generally speaking, fixation count and fixation duration are positively correlated with search time¹⁵. In other words, they are negatively correlated with search efficiency. Fixation duration in the search task is also related to the number of items appearing on the interface²⁵. Scan path is the sequence of fixations, which can be seen as the trajectory with points and lines. Scan path can reflect the user's eye movement mode^21,22. How to determine the next fixation position to accomplish the search operation more efficiently has been a question for many researchers²⁶. A previous study found that when the user searches the interface, the next search direction is reselected, and the previous fixation is irrelevant²⁷. However, when searching for seven items in a circular array, users generated about a quarter of the return eye jump²⁸, which means that users may have an adjustment process before they find the next target. Some theories suggest that the next fixation is more likely to move to the location that users believes is most relevant to the search task^29–33. Some research believe that fixation will move between two potential targets, which will maximize the efficiency of search³⁴. A more complex theory suggests that the best location for the next fixation is determined by planning. The current situation and the potential benefits in the future should be considered. And other studies verified that eye movements are indeed planned according to probabilistic planning^26,35.

How to compare scan paths has also attracted the attention of many scholars. A study compared the shape of scan path in different modes, but simply did a subjective assessment of the scan path plots without using statistical data¹⁷. Another method is to use vectors to represent scan path³⁶. A method called MultiMatch was used to compare scan paths in multiple dimensions³⁷. A few studies have also looked at the angle of the scan path. To evaluate search tasks, an analysis method known as "Eight Point Star" (8PS) was used³⁸. In this model, the angle refers to the difference between the actual and the expected angle of the paths. In a simple visual search task, the model is applicable and can effectively summarize the angle of scan path for analysis. However, in complex search tasks, the connection line between two target items is often complex. Another method called long common sequence (LCS) was used to count the same paths between the actual scan path and the best scanpath³⁹.

The values of angles between vectors should be considered as well, and one study divided them into eight intervals²². The percentage of angles near to 0° and 360° is higher than other angles, according to this study, and it varies depending on the types of tasks. We found that the angular distribution in this model has obvious symmetry. Therefore, we create a new model that divides all angles into four intervals. The angle in this model is formed by two successive lines in scan path, which is related to three fixations: the previous one, the current one, and the next one. This includes previous decision, current status, and future planning. Because of this, the angle (called scan-path angle in this study) may help us better understand the strategy of user's eye movement in search tasks.

This research aims to compare the impact of changes in distraction and layout on search efficiency and the relationship between them and search strategy. In addition to fixation count, fixation duration and saccade count, more complex scan paths were also considered. However, it is difficult to find the strategy of how to select the next fixation location only by analyzing fixation, saccade, and scan path, or just observing the trajectory of eye movement. Therefore, a new method is needed to explore the relationship between fixations. The trajectory of eye movement can be studied from a deeper level: fixations are connected in sequence to form angles, which is related to the previous fixation, the current fixation, and the next fixation, and represents the direction of the current eye movement.

There were two experiments. Both experiments required participants to look for targets. The first experiment was the distraction group and the second experiment was the layout group. The study was approved by the Qilu University of Technology Ethics Committee. All experiments followed the principles of the Declaration of Helsinki and informed consent was obtained from all subjects.

Experiment 1

Forty-three participants (16 females, age:18.98 ± 0.98) participated in the study. All of the participants were undergraduates who were recruited from universities. Before the experiment, all the participants had rich experience in using graphical interface and computer. Participants were Chinese characters users and were familiar with the order of the English alphabet. In order to exclude the influence of poor vision on the experimental results, participants were tested before the experiment to ensure that participants had normal or corrected-to-normal vision.

The page was designed with 36 squares, 18 of which have capital letters. There were two types of tasks: non-distraction and distraction (Fig. 1), in which all items were arranged in a pseudo-random way into 36 squares. Participants were asked to face the monitor and click on the squares in the order of alphabet. A trial ended when all 18 letters were correctly clicked, and the following trial appeared.

Before entering the laboratory, the participants were tested for familiarity with the English alphabet. Then they practiced four times, twice with non-distraction and twice with distraction. The whole experiment lasted about 10 to 15 minutes (Fig. 2). Before the start of non-distraction and distraction, a "+" lasting 1000 milliseconds was presented in the center of the screen, and the participants were required to look at the location of "+". Participants need to finish five trails in non-distraction and five trails in distraction. After the experiment, participants were asked to complete a 7-piont subjective scale to score the difficulty of non-distraction and distraction (1 means very simple, and 7 means very difficult).

Experiment 2

The data were collected from 36 participants (14 females), aged between 18 and 27, with an average age of 21.44 (SD = 1.83). The participants were undergraduates and postgraduates in the College of Art and Design. The participants had rich experience in using graphical interface and online shopping website. All participants had normal or corrected-to-normal vision.

There were four words on the left side of the monitor and four groups of pictures of 16 commodities on the right side. Each group had one picture corresponding to one word on the left side. The participants were asked to observe the words on the left side and find the corresponding goods on the right side. Once the participants found four corresponding commodities, they pressed the space bar to enter the next trial. Participants were asked to complete three groups of tasks according to the layout: horizontal, vertical, and square (Fig. 3).

Before the formal experiment, the participants were given three exercises of horizontal, vertical and square. The participants had to complete three groups of 30 trials (Fig. 4). Each task began with a "+" that lasted 2000 milliseconds and required the participants to stare at the position of the "+." In order to eliminate the influence of learning effect, the experiment adopted Latin square design. Experiment 2 lasted about 5 to 10 minutes. After the experiment, participants were asked to complete a 7-piont subjective scale as the same as in Experiment 1.

Data collection

The evaluation data were from subjective questionnaires. Eye movement data were obtained from the eye tracker named Tobii Pro glasses 2. Attention lasting more than 200 milliseconds, and no more than two degrees in diameter, can be regarded as fixation. The coordinate information of all fixations was collected. In order to better summarize the scan path, this study established a new model, called Scan-path Angle Model (SAM), as shown in Fig. 5. The value of the angle can be calculated using the fixation coordinates. In SAM, all angles are divided into four intervals according to the value: SA1(0°,45°], SA2(45°, 90°], SA3(90°, 135°] and SA4(135°, 180°]. According to the value, the angle belongs to SA1, SA2, SA3 or SA4. To better analyze scan-path angle, we calculated their percentages. SA1P is equal to the number of angles in SA1 divided by the number of all angles.SA2P, SA3P and SA4P were obtained by the same method.

Search time and subjective evaluation

Distraction increased search time. The result of matched samples t-test showed a significant difference (t=-6.721, p < 0.001) of subjects’ search time in non-distraction (M = 117.43, SD = 21.837) and distraction (M = 144.58, SD = 31.583). Meanwhile, the change of layout also increased the search time. The difference [F (2,70) = 11.045, p < 0.001, η² = 0.240] of subjects' search time in horizontal (M = 102.86, SD = 34.42), vertical (M = 107.08, SD = 40.60) and square (M = 117.53, SD = 40.12) was also significant, which was tested by one-way repeated-measures ANOVA. The result was the same as we expected. Distraction and layout changed the difficulty, so the search efficiency was also different.

The difference (t=-22.405, p < 0.001) of the score of difficulty evaluation in distraction (M = 2.93, SD = 1.01) and distraction (M = 5.23, SD = 1.00) was significant. The results in horizontal (M = 3.00, SD = 1.01), vertical (M = 3.42, SD = 1.40) and square (M = 3.86, SD = 1.29) were also significantly different [F(1.418, 49.626) = 10.313, p = 0.001, η² = 0.228]. The results showed that the participants' subjective evaluation of difficulty was consistent with the objective difficulty of the task, which was reflected in the search time.

Eye-tracking metrics

When there were distractions in the page, fixation count, fixation duration and saccade count all increased (Fig. 6). The result of matched samples t-test showed that fixation count, fixation duration and saccade count all had significant effects (p < 0.05) in non-distraction and distraction. The change of layout also affected fixation count, fixation duration and saccade count. However, only fixation count had significant effects [F (2,70) = 23.556, P < 0.001, η² = 0.402]. More specific trends are denoted by multiple comparisons of fixation count, fixation duration and saccade count, as shown in Fig. 7. The correlation analyses showed that fixation count, fixation duration and saccade count have significant effect (p < 0.05) on search time in non-distraction and distraction (Fig. 8) and in horizontal, vertical, and square (Fig. 9). All eye-tracking metrics were positively correlated with search time. The above results showed that the difficulty caused by the change of distraction and layout increased, making participants need more eye movement and more observing time.

Scan path

The fixation points were connected by lines sequentially to obtain the trajectory of fixations which represents scan path. Both distraction and layout affected the trajectory of participants' fixation. Compared with non-distraction, distraction made participants' scan paths denser, but the random distribution of fixations did not change (Fig. 10). The difference of layout did not change the density of scan path, but the shape changed. Accumulated horizontal lines can be seen in horizontal, accumulated vertical lines can be seen in vertical, while there were more short lines in square (Fig. 11).

We have tried to create SAM (Scan -path Angle Mode) and apply it to the analysis. The proportional distribution of scan-path angle was a surprising discovery. The results of SA1P were very close in non-distraction (M = 32.72%, SD = 0.0208) and distraction (M = 31.66%, SD = 0.0503), and larger than SA2P, SA3P and SA4P, as shown in Fig. 12. This trend was more obvious in layout group, in which the results of SA1P in horizontal (M = 45.00%, SD = 0.0535), vertical (M = 41.05%, SD = 0.0499) and square (M = 44.39%, SD = 0.0402) were shown in Fig. 13. When distraction added, the results of SA2P, SA3P and SA4P were very similar in non-distraction and distraction. However, after the layout changed, the results of SA2P, SA3P and SA4P were different in horizontal, vertical, and square.

Search time was influenced by both distraction and layout. When the goals of search tasks are the same, a rise in search time indicates a higher task difficulty and a lower search efficiency. Distraction increases the number of items and the complexity of the page, so the search task is more difficult^12,15,17. The search difficulty of the square was the highest, followed by the vertical and then the horizontal. The different layout affects the way of eye movement, which is related to people's habit of searching visual information in the past²⁵ and affect search efficiency. The difficulty indicated in the search time was consistent with the subjective evaluation results of participants on the difficulty. Search behaviors, notably eye movement, influenced participants' perceptions of difficulty. Each eye movement and overall movement mode has an impact on search efficiency, which in turn has an impact on difficulty evaluation^15,40,41.

The results of the three eye-tracking metrics were significantly influenced by distraction. Fixation count, fixation duration, and saccade count all increase as difficulty increases, which is consistent with previous researches^3,8,15. The influence of layout is eliminated in non-distraction and distraction tasks since all elements are arranged pseudo-randomly. Distraction resulted in three major changes: the number of items rose, the variety of items increased, and the density of information increased. As a result, the value of target divided by all items has decreased; the interface's clutter has increased; and the target has become less visible against the cluttered background. Therefore, additional fixations and saccades are required to discriminate targets and incorrect items. Simultaneously, the increase in items and clutter caused participants to do redundant scans in the same area, which was ineffective. The effects in layout on eye-tracking metrics were not as obvious as those produced by distraction. These three metrics cannot explain the reasons for the change in search difficulty caused by layouts. The number and type of items in these three layout tasks did not change, as did the density of information. Only the arrangement of the items, that is, the positional connection between items, changed. Therefore, not only should each individual eye movement be evaluated, but also the positional relationship between them^19,37,38.

Trajectory plots naturally indicated the movement mode of participants' fixation^17,21. Although the quantity and density of fixations were different in non-distraction and distraction, the outer contours of the two plots formed of scan path were similar. In both plots, the distribution of fixations was also random. This random arrangement was similar to the way items were arranged in the search tasks. It can be inferred that the density and arrangement of fixations in the trajectory plots were influenced by the items in the search task. The density of fixation in the trajectory plots did not differ significantly between the three tasks with various layouts. The number and density of items on each page were the same in all three tasks, demonstrating once again that the intensity of fixation in the trajectory plots corresponded to the items on the page. The shapes of scan paths were clearly different. The shapes of scan path matched the layout of the pages. This finding implied that the change in layout affected participants' fixation movement mode. This is quite understandable. When the task goal is clear in search, the bottom-up approach often plays a leading role⁴². Therefore, the direction of the next fixation was guided by the layout²⁶, so the scan path formed a similar layout. Horizontal and vertical are the common arrangement of information, especially text. Horizontal eye movement delivered higher efficiency, which may be related to people's long-term reading habits^22,25. The square is not common, and this unfamiliar layout made participants spend more time^43,44.

We separated all scan-path angles into four intervals and calculated the percentage of each interval, which had never been done before. SA1P was the largest in both the non-distraction and distraction tasks, and the two values were quite close. Furthermore, the percentages of the other three intervals were similar as well. Distraction increased the difficulty of searching in three ways, but participants' search strategies remained unchanged, and they preferred to adjust eye movement with a small angle. Therefore, the number of items, the type of items and the density of the page cannot change the angle strategy, and difficulty is not a sufficient condition to change the scan-path angle. The scan-path angles less than 45 degrees were still the greatest and the percentages were extremely close in horizontal, vertical, and square tasks when the layout was changed, which was similar with the distraction group's results. The results of the other three angle intervals, on the other hand, varied depending on the layout. This result corresponds to the scan path: the shape has changed, and this change is mainly caused by angles of more than 45 degrees. Combined with the results of the distraction group, choosing a small angle is a basic strategy, no matter what the search task is. Interface design is unlikely to have an impact on this tendency. This tendency is related to reading habits^22,25 and eye discomfort. Common visual information is arranged horizontally or vertically. Guided by such layouts, the angle of eye movement is mostly small. Varied types of eye movement cause different levels of tiredness and comfort in persons, and simple angle adjustments can lessen the physiological strain.

Distraction makes search task more difficult, and search time is positively correlated with fixation count, fixation duration, and saccade count. When the layout of the task changes, the difficulty of the task varies as well, but only the fixation has a significant difference. Distraction cannot change the shape of scan path, but can change the density of fixation in the scan path. The change of layout will not change the density of fixation, while the shape of scan path is related to the layout of the page of search task. In all cases, scan-path angles less than 45 degrees are always the most.

Author contributions

Y.M. conceived the basic idea and validated the concept of operation through analytical and numerical modelling, designed and optimized the device, analyzed the results and wrote the paper. Z.W helped in experiments and data process. G.R. helped in writing and revising the manuscript. All the authors contributed to the general discussion and revision of the manuscript.

Data availability

The datasets used and/or analysed during the current study available from Yongchun Mao on reasonable request.

Treisman, A. M. & Gelade, G. A feature-integration theory of attention. Cognitive Psychology 12, 97–136 (1980).
Young, A. H. & Hulleman, J. Eye Movements Reveal how Task Difficulty Moulds Visual Search. Journal of Experimental Psychology Human Perception & Performance 39, 168–190 (2013).
Wu, X., Gedeon, T., Xue, C., Zhang, W. & Wang, L. Comparison of the Congruent Effects on Fixation/Saccade and Pupil Dilation Influencing Information Feature Searching. Jisuanji Fuzhu Sheji Yu Tuxingxue Xuebao/Journal of Computer-Aided Design and Computer Graphics 31, 1636–1644 (2019).
Patel, G. A. & Sathian, K. Visual search: Bottom-up or top-down? Frontiers in Bioence 5, D169-93 (2000).
Hou, G. & Hu, Y. Designing Combinations of Pictogram and Text Size for Icons: Effects of Text Size, Pictogram Size, and Familiarity on Older Adults’ Visual Search Performance. Hum Factors 00187208211061938 (2021) doi:10.1177/00187208211061938.
Duncan, J. & Humphreys, G. W. Visual search and stimulus similarity. Psychological Review 96, 433 (1989).
Gong, Y., Zhang, S. Y., Liu, Z. F. & Shen, F. Eye movement study on color effects to icon visual search efficiency. Journal of Zhejiang University (2016).
Pang, L., Fan, Y., Deng, Y., Wang, X. & Wang, T. Mental Workload Classification By Eye Movements In Visual Search Tasks. in 2020 13th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI) (2020).
Nei De R, M. B. & Zelinsky, G. J. Cutting through the clutter: Searching for targets in evolving complex scenes. Journal of Vision 11, 7–7 (2011).
Orth, U. R., Wirtz, J. & Mckinney, A. Shopping experiences in visually complex environments: a self-regulation account. Journal of Service Management (2009).
Wang, Q., Yang, S., Liu, M., Cao, Z. & Ma, Q. An eye-tracking study of website complexity from cognitive load perspective. Decision Support Systems 62, 1–10 (2014).
Wu, L., Zhu, Z., Cao, H. & Li, B. Influence of information overload on operator’s user experience of human–machine interface in LED manufacturing systems. Cognition Technology & Work 18, 161–173 (2016).
Benjamins, J. S., Hooge, I. T. C., Elst, J. C. van, Wertheim, A. H. & Verstraten, F. A. J. Search time critically depends on irrelevant subset size in visual search. Vision Research (2009).
Lamy, D. & Zoaris, L. Task-irrelevant stimulus salience affects visual search. Vision research 49, 1472–1480 (2009).
Liu, K. & Or, K. L. Effects of Image Segmentation and Target Number on Visual Search: An Eye-Tracking Study. Proceedings of the Human Factors & Ergonomics Society Annual Meeting 62, 1987–1991 (2018).
Dianat, I., Adeli, P., Jafarabadi, M. A. & Karimi, M. A. User-centred web design, usability and user satisfaction: The case of online banking websites in Iran. Applied Ergonomics 81, 102892- (2019).
Guo, F., Chen, J., Li, M., Wei, L. & Zhang, J. Effects of visual complexity on user search behavior and satisfaction: an eye-tracking study of mobile news apps. Universal Access in the Information Society (2021).
Li, M., Zhong, N. & Lü, S.-F. Exploring visual search and browsing strategies on web pages using the eye-tracking. Beijing Gongye Daxue Xuebao/Journal of Beijing University of Technology 37, 773–779 (2011).
Falschlunger, L., Eisl, C., Losbichler, H. & Grabmann, E. Report optimization using visual search strategies: An experimental study with eye tracking technology. in 209–218 (2015). doi:10.5220/0005251702090218.
Li, M. et al. Strategy and processing mode of visual search under information overload on Web pages: An eye-tracking study. Beijing Gongye Daxue Xuebao/Journal of Beijing University of Technology 38, 390–395 (2012).
Wooding, D. S. Fixation maps: Quantifying eye-movement traces. in Proceedings of the Eye Tracking Research & Application Symposium, ETRA 2002, New Orleans, Louisiana, USA, March 25–27, 2002 (2002).
Amor, T. A., Reis, S., Campos, D., Herrmann, H. J. & Andrade, J. S. Persistence in eye movement during visual search. Scientific Reports.
Bruin, J. A. D., Malan, K. M. & Eloff, J. H. P. Saccade deviation indicators for automated eye tracking analysis. in Proceedings of the 2013 Conference on Eye Tracking South Africa (2013).
Guo, F., Ding, Y., Liu, W., Liu, C. & Zhang, X. Can eye-tracking data be measured to assess product design?: Visual attention mechanism should be considered. International Journal of Industrial Ergonomics 53, 229–235 (2016).
Ojanp, H., N?S?Nen, R. & Kojo, I. Eye movements in the visual search of word lists. Vision Research 42, 1499–1512 (2002).
Hoppe, D. & Rothkopf, C. A. Multi-step planning of eye movements in visual search. Scientific Reports 9, (2019).
Findlay, J. M., Brown, V. & Gilchrist, I. D. Saccade target selection in visual search: the effect of information from the previous fixation. Vision Research 41, 87–95 (2001).
Hooge, I. & Erkelens, C. J. Control of fixation duration in a simple search task. 58, 969–976 (1996).
Itti, L. & Koch, C. A saliency-based search mechanism for overt and covert shifts of visual attention. Vision Research 40, 1489–1506 (2000).
Itti, L. & Baldi, P. Bayesian surprise attracts human attention. Vision Research 49, 1295–1306 (2009).
Renninger, L. W., Coughlan, J., Verghese, P. & Malik, J. An information maximization model of eye movements. Advances in neural information processing systems 17, 1121–1128 (2005).
Renninger, L. W., Verghese, P. & Coughlan, J. Where to look next? Eye movements reduce local uncertainty. J Vis 7, 6 (2007).
Najemnik, J. & Geisler, W. S. Eye movement statistics in humans are consistent with an optimal search strategy. Journal of Vision 8, (2008).
Najemnik, J. & Geisler, W. Optimal eye movement strategies in visual search. Nature 139, 1152–1153 (2005).
Kaelbling, L. P., Littman, M. L. & Cassandra, A. R. Planning and acting in partially observable stochastic domains. Artificial Intelligence (1998).
Jarodzka, H., Holmqvist, K. & Nyström, M. A vector-based, multidimensional scanpath similarity measure. in Symposium on Eye-tracking Research & Applications (2010).
Richard et al. It depends on how you look at it: Scanpath comparison in multiple dimensions with MultiMatch, a vector-based approach. Behavior Research Methods 44, 1079–1100 (2012).
Veneri, G. et al. Evaluating the human ongoing visual search performance by eye tracking application and sequencing tests. Comput Methods Programs Biomed 107, 468–477 (2012).
Richard, D. et al. How task demands influence scanpath similarity in a sequential number-search task. Vision Research 149, 9–23 (2018).
Murata, A., Takahashi, Y. & Furukawa, N. Evaluation of display design based on human’s eye movement. in vol. 2 587–592 (2002).
Liu, Z., Xu, Z., Zhou, Q., Xie, F. & Zhou, S. A study of multi-target visual search by eye movement behavior. in vol. 9174 34–44 (2015).
Hwang, A. D., Higgins, E. C. & Pomplun, M. A model of top-down attentional control during visual search in complex scenes. Journal of Vision 9, (2009).
Stefani, M., Sauter, M. & Mack, W. Delayed disengagement from irrelevant fixation items: Is it generally functional? Attention, Perception, and Psychophysics 82, 637–654 (2020).
Tzanidou, E., Minocha, S., Petre, M. & Grayson, A. Revisiting Web design guidelines by exploring users’ expectations, preferences and visual search behaviour. in 421–438 (2006). doi:10.1007/1-84628-249-7_27.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Efficiency and strategy of visual search: a study on eye movement and scan path

Status:

Version 1

Abstract

Figures

Introduction

Methods

Results

Discussion

Conclusion

Declarations

References

Additional Declarations

Status:

Version 1