The 100 Most Cited Papers Concerning the Insular Cortex of the Brain: A Bibliometric Analysis

Background: The insula is one of the most researched brain regions with many highly cited papers. However, unlike the literature of other fields, there is currently no study that has identified the 100 most cited papers within the literature of the insula. The aim of the current study was to fill in the knowledge gap by determining which publications concerning the insula have been cited most often, who contributed to them, and what topics they were dealing with. Methods: The Web of Science online database was searched to identify the 100 most cited publications mentioning the insular cortex in their titles, abstracts or keywords. To systematically exclude irrelevant publications, the search strategy was finalized as: TS = (insula OR insular OR “island of Reil”) NOT TS = (“insular biogeography” OR “insular mammal*” OR “*insular lymphatic*”) NOT WC = (“Geochemistry and Geophysics” OR “Ecology”). The identified publications were sorted in descending order of citation count. The 100 most cited publications concerning the insula of the brain were identified and their bibliometric data was extracted and assessed. The VOSviewer software was used with default parameters to generate a bubble map that analyzes and visualizes the words/phrases used in the titles and abstracts of the publications. Results: There were 67 articles on experiments/lab studies and 33 meta-analyses/reviews but no opinion or methods paper. They had an average of 943.4 citations (or 62.9 citations per year), 93.5 references and 13.4 pages. There were 35 papers published in open access. USA was the major contributing country. The most top-ranked publications were concerning emotion, salience and pain. Conclusion: Two-thirds of the publications concerned the normal brain function/mechanism (n = 67), whereas 20 publications concerned disease/therapeutic intervention and another 13 concerned normal anatomy. For the 67 original articles, 57 used human subjects whereas 10 used animal models. MRI was the commonest modality (n = 37), followed by PET (n = 16). Nine articles investigated by histology, two by multiple modalities and three by other modality.


INTRODUCTION
The word "insula" was firstly coined by Professor Reil and the insula is also known as "Island of Reil" (Reil, 2002;Binder et al., 2007). It is a fascinating structure of the cerebral cortex that has multiple functions and clinical relevance. For instance, it is involved in sensory processing (Naghavi et al., 2007;Yeung et al., 2016Yeung et al., , 2018a, motor control (Ackermann and Riecker, 2004), interoceptive awareness (Naqvi and Bechara, 2010), autonomic control (Yasui et al., 1991), and social emotions (Lamm and Singer, 2010). It is often investigated in clinical topics such as addiction (Naqvi and Bechara, 2009), anorexia nervosa (Wagner et al., 2008), and stroke (Cereda et al., 2002). In short, the insula is involved in many researches, ranging from neurology, neuropathology, psychiatry, and psychology.
The evaluation of the academic impact of the published literature in terms of citation count has been gaining attention. It not only allows the evaluation of research performance, but also helps investigators refine their research blueprint based on the research trends and overall landscape of the research field (Brinjikji et al., 2013;Yeung et al., 2017a). There are already numerous reports that identified the 100 most cited articles, also known as citation classics, in various fields such as imaging (Brinjikji et al., 2013), radiology (Yoon et al., 2013;Pagni et al., 2014), neuroimaging (Kim et al., 2016), neurosurgery (Ponce and Lozano, 2010), and neurosciences (Yeung et al., 2017a). Analyses based on citation counts are primarily supported by an earlier notion that citation counts could be correlated to peer judgments (Bayer and Folger, 1966;Garfield, 1970), which are in turn commonly accepted as a way to rank scientific performance (Garfield, 1979). Citation counts are not without problems. For instance, MacRoberts and MacRoberts (1989) have pointed out that potential problems of citation analysis included inability to assess formal and informal influences not expressed as journal citations, biased citation behavior, self-citation, different meaning of each citation, variations in citation rates due to various background parameters of a publication, and technical shortcomings of citation databases. Nonetheless, citation analysis and journal impact factors are used in many circumstances, such as determining departmental funding in the universities, assessing research grant proposals, and deciding promotions and appointments of academic staff (Adam, 2002). Therefore, results reported in bibliometric studies enabled researchers to quickly identify the most cited publications of the respective field so that they might further the research works or develop new research directions based on these cornerstones (Khan et al., 2017). However, such bibliometric evaluation has not been applied to the research of the insula, a versatile multifunctional brain region and a research hot topic with highly cited papers (Yeung et al., 2017c;Yeung, 2018). Therefore, the aim of the current study was to fill in the knowledge gap by determining which publications concerning the insula have been cited most often, who contributed to them, and what topics they were dealing with.
In particular, the current study tried to analyze what factors were related to the citation counts, e.g., normalized citation count, reference count, usage count, years since publication, number of authors, page count, whether the publications were in open access, the publication type, study topic, the study model, and the modality used.

Search Method and Strategy
The search was performed on 18 July 2018 using the Web of Science (WoS) online database hosted by Clarivate Analytics. Based on a pilot search, the author found that some of the highly cited publications mentioning the word "insula" or its derivatives are not related to the insular cortex of the brain, such as those concerning insular biogeography (investigation of animals and plants in isolated natural environments) and those concerning peri-insular lymphatics of the pancreas. To systematically exclude these irrelevant publications instead of having a subjective manual exclusion, the search strategy was finalized as: TS = (insula OR insular OR "island of Reil") NOT TS = ("insular biogeography" OR "insular mammal * " OR " * insular lymphatic * ") NOT WC = ("Geochemistry and Geophysics" OR "Ecology"). It identified publications mentioning the insula in their titles, abstracts, author keywords or KeyWord Plus (keywords indexed by WoS). Keywords serve the function of providing rapid access to scientific works (Soos et al., 2013) and are useful for bibliometric analyses (Stock and Stock, 2013;Zhang et al., 2016;Vargas-Quesada et al., 2017), and thus were considered together with abstracts and titles. No other restriction was placed on the search, such as publication year. The identified publications were sorted in descending order of citation count. The 100 most cited publications concerning the insula of the brain were identified and their bibliometric data was extracted and assessed, including the author list, affiliation, country of affiliation, journal, language, WoS category of the publication, publication year, number of references, number of citations, number of pages of the manuscript, and status of open access, the publication type (experiment/lab studies, methods, meta-analysis/review/opinion), and the study topic (normal brain function/mechanism, disease/therapeutic intervention, normal anatomy). For experiment/lab studies, the study model (human, animal, computational, multiple) and modality [magnetic resonance imaging (MRI), positron emission tomography (PET), multiple, histology, others] were also assessed.

Network Visualization
VOSviewer, software designed to visualize bibliometric data (Van Eck and Waltman, 2009), was used to generate a bubble map that analyzes and visualizes the words/phrases used in the titles and abstracts of the 100 publications. All words/phrases that appeared in two or more of the 100 publications were included. Binary counting was used, implying that multiple appearances in a single publication counted as one. Default visualization parameters were chosen. Two bubbles were in closer proximity if they co-appeared in a larger number of publications. The number of publications in which the word/phrase appeared determined the bubble size. The bubble color was determined by the averaged citations per publication containing the word/phrase. This method has been successfully used to analyze the titles and abstracts of publications from neuroimaging, neurosciences, food sciences and pharmacology (Yeung et al., 2017b(Yeung et al., ,c, 2018bYeung, 2018).

Statistical Analysis
Exploratory analysis was performed to evaluate if the citation count had significant associations with the numbers of normalized citation count (citations per year after publication), years since publication, references, pages, authors, as well as the usage count since 2013 (i.e., the number of times the full text of a publication was accessed or saved through WoS), status of open access, publication type, study topic, study model, and modality. Pearson's correlation tests were conducted to assess the former six variables, whereas two-sample t test was conducted to assess the latter five variables. For the last four variables, either only two groups had values or the groupings were binarized (e.g., study topic binarized into findings from healthy brains/patients and from patients with diseases/disorders; and modality binarized into MRI and non-MRI). The statistical analysis was performed with SPSS 25.0 (IBM, New York, USA). Test results with p < 0.05 were considered statistically significant.

The Overview
The publication years of top 100 most cited publications of the insula were happened to be from 1982 to 2011. The list was presented in Table 1. There were 67 articles on experiments/lab studies and 33 meta-analyses/reviews but no opinion or methods paper. They have received a total of 94,340 citations from 61,624 citing articles, equaling an average of 943.4 citations (SD = 506.2, range = 544-2924). When the citation counts were normalized by the number of years since publication, the average value became 62.9 (SD = 46.3, range = 17.9-268.7). The usage count since 2013 through WoS had an average of 140 times (SD = 143, range = 6-680). On average they had a whopping number of 93.5 references (SD = 73.4, range = 19-387). The average length of the manuscript was 13.4 pages (SD = 8.3, range = 3-47). There were 35 papers published in open access (gold, bronze or green open access). All of them were written in English. Two-thirds of the publications concerned the normal brain function/mechanism (n = 67), whereas 20 publications concerned disease/therapeutic intervention and another 13 concerned normal anatomy.
For the 67 original articles, 57 used human subjects whereas 10 used animal models. MRI was the commonest modality (n = 37), followed by PET (n = 16). Nine articles investigated by histology, two by multiple modalities and three by other modality.

Major Contributors
The citation classics were contributed by 361 authors working in 130 affiliations in 13 countries. They were published in 38 journals. The five major contributing affiliations were Stanford University (n = 8), Harvard University (n = 8), University College London (n = 6), Massachusetts General Hospital, National Institute of Mental Health, University of Iowa and Washington University (each n = 5). The top five major contributing countries were USA (n = 70), UK (n = 27), Canada (n = 11), Germany (n = 6), and Italy (n = 5). The top five major contributing journals were Journal of Neuroscience (n = 8), Journal of Comparative Neurology (n = 7), Biological Psychiatry, NeuroImage, and Neuron (each n = 6). The top five major WoS categories to which the citation classics were assigned were neurosciences (n = 79), multidisciplinary sciences (n = 14), clinical neurology (n = 13), psychiatry (n = 13), and neuroimaging, radiology/ nuclear medicine/ medical imaging and zoology (each n = 7). Note that each publication could be assigned to multiple WoS categories. Figure 1 shows the bubble map that visualizes the words/phrases used in the titles and abstracts of the 100 publications. Various aspects of the publications might be observed. First, functional magnetic resonance imaging (appearances = 24, average citations = 937.2) and positron emission tomography (appearances = 15, average citations = 944.1) were the two main imaging modalities. When the underlying data of the bubble map was examined, computed tomography, the popular modality used in medical field, was absent. This has implied that the most cited publications on the insula have been focusing on functional neuroimaging heavily.

DISCUSSION
The current study has revealed that the citation classics concerning the insula were published across a long period of time spanning nearly four decades. Most of them were original articles and contributed by authors working in affiliations based in USA. Moreover, most of them were classified by WoS as dealing with neurosciences, and some of them as clinical neurology, psychiatry and neuroimaging. However, it should be noticed that seven of them were classified as dealing with zoology. This was due to the fact that Journal of Comparative Neurology encourages submissions that compare brain functions among species, also known as systems neuroscience, that can be considered as relevant to zoology. The large share of citation FIGURE 1 | Bubble map visualizing the words/phrases used in the titles and abstracts of the 100 most cited publications concerning the insular cortex of the brain. All words/phrases that appeared in two or more of the 100 publications were included. Binary counting was used, implying that multiple appearances in a single publication counted as one. Default visualization parameters of the software VOSviewer were chosen. Two bubbles were in closer proximity if they co-appeared in a larger number of publications. The number of publications in which the word/phrase appeared determined the bubble size. The bubble color was determined by the averaged citations per publication containing the word/phrase. classics contributed by Journal of Comparative Neurology reminds us the importance and usefulness of investigations using different species as models, such as monkeys and rats. Though all of the publications were dealing with the insula, it could be observed that brain regions often mentioned together with the insula included the amygdala, thalamus, anterior cingulate and prefrontal cortex, cerebellum, and hippocampus. For instance, the top-ranked most cited insula publication investigated gray matter volume loss in aging adults, and reported that the insula and cingulate were affected but not the amygdala and the hippocampus (Good et al., 2001). Meanwhile, the 2nd-ranked most cited insula publication reported that the anterior cingulate and orbital frontoinsular cortices were involved in a salience network that linked to limbic and subcortical regions, and an executive network that communicated with prefrontal and parietal cortices (Seeley et al., 2007). A previous bibliometric report has also pointed out that neuroimaging studies on food stimulations generally had more citations if the insula, orbitofrontal cortex and amygdala were involved (Yeung, 2018). These findings have reflected that notion that the insula is an integration hub that connects multiple brain regions. Besides, Table 1 has shown that the most top-ranked publications were concerning emotion, salience and pain. One possible reason for articles investigating normal brain functions/mechanisms from healthy subjects having more citations than those investigating patients was that findings from normal subjects might be more generalizable and applicable to a larger population.
The positive correlation between citation count and reference count is consistent to many previous studies (Peters and Van Raan, 1994;Lokker et al., 2008;Farshad et al., 2013;Hanssen and Jørgensen, 2014;Fox et al., 2016). However, the Pearson correlation coefficient was very small (0.289). It implied that if a larger dataset or a different dataset was examined, it might become insignificant. The non-correlation between citation count and year since publication is also consistent with previous findings (Hanel and Haase, 2017). The current sample was small and not random, but the results of such exploratory analysis seemed to suggest that the selected insula literature did not deviate much from other representative body of literature with regards to influencing factors of citation count. The noncorrelation between citation count and author count, page count or open access status, however, are inconclusive because previous studies have reported mixed results (see Table 1 of Hanel and Haase, 2017).
The most important limitation of the current study was that bibliometric analysis cannot assess the validity of or the level of scientific evidence reported by the analyzed publications. A highly cited article may not necessarily have high scientific quality. Moreover, the citation count received by a publication might depend on other factors that might not be fully considered in the current study. To start with, an article may be cited more if it was published in a more visible journal (Brink, 2013); For example, articles published in Chinese scientific journals tended to have few citations and thus leading to low journal impact factors (Ren and Rousseau, 2002). Meanwhile, the study model might also be an influencing factor. It was reported that basic research generally received more citations than clinical studies (Van Eck et al., 2013); and methods papers were often cited as a common practice so they accumulated many citations (Van Noorden et al., 2014). In the current study, human studies on average had 179 more citations than animal studies. The imaging modality may also be a factor, as a previous bibliometric study has reported that most of the highly cited neuroimaging papers used MRI compared to PET, computed tomography and ultrasonography (Kim et al., 2016). The current study also found that MRI was the commonest modality, but its usage did not significant influence citation count. Finally, there could be authors who self-cited a lot, which might then lead to an increased visibility of the articles and hence more citations from others (Fowler and Aksnes, 2007). Readers should be clearly aware of these confounding factors when they interpret the results from the current manuscript. Also, it is intuitive to reason that the citation counts from older publications should usually be larger than more recent publications, thus making the ranking and analyses biased. Nonetheless, the current results have demonstrated a strong positive correlation between total and normalized citation counts. Maybe it was partly due to the fact that the latest publication among the most cited 100 publications has already been published for 7 years. Such a long period of time might have already allowed publications to cumulate a large number of citations, as it was reported that the number of annual citations usually tended to reach a maximum of plateau 2-3 years after publication (Hansen and Henrikson, 1997).
To conclude, the current study has identified and analyzed the 100 most cited publications concerning the insula of the brain. These publications have shown that the insula is a multifunctional brain region. The top ranked papers on the list tended to focus on emotion, salience and pain. Brain regions often mentioned together with the insula included the amygdala, thalamus, anterior cingulate and prefrontal cortex, cerebellum, and hippocampus. Citation count of these 100 highly cited publications correlated with their normalized citation count, number of references and usage count in WoS, but not years since publication, number of authors, page count and status of open access.

AUTHOR CONTRIBUTIONS
The author confirms being the sole contributor of this work and approved it for publication.