Furthermore, interdisciplinarity and cross-disciplinarity have been buzzwords for the last few years, which are used to describe contributions from and collaborations among several or more disciplines. Interdisciplinary means that the content of research is not only a method or ability in a field but a field that involves more [20 (link)]. Through interdisciplinary research, we can more comprehensively understand the research content of a field. Interdisciplinary inevitably exists between disciplines, indicating that the scope involved in a certain field is constantly expanding [21 (link)]. Meanwhile, research areas constitute a subject categorization scheme that is shared by all Web of Science product databases. The literature indexed by WoSCC is assigned to at least 1 subject category, which is mapped to 1 research area [22 ]. VOSviewer—a software tool developed by Nees Jan van Eck and Ludo Waltman at Leiden University's Centre for Science and Technology Studies [23 (link)]—was employed to visualize the interdisciplinary collaboration on the basis of subject categorization of publication [24 (link)]. Each node represents a discipline, whereas the connection between nodes represents collaborations between disciplines. In addition, nodes with a close connection are assigned the same color to form their respective clusters. Furthermore, a co-occurrence matrix was generated by using the Bibliographic Item Co-occurrence Mining System (BICOMS) [25 (link)] to calculate the centrality, which includes degree centrality, closeness centrality, and betweenness centrality by using Ucinet6.6 [26 (link)]. Degree centrality is simply the number of tie of a given type that a node has; closeness is an inverse measure of centrality in the sense that large numbers indicate that a node is highly peripheral, whereas small numbers indicate that a node is more central; betweenness centrality is a measure of how often a given node falls along the shortest path between 2 other nodes [27 ]. Moreover, we analyzed the centrality in the different periods of time based on the top 5 centralities over the period from 1997 to 2017.
In addition, we used Cortext to visualize the evolution of individual disciplines and interdisciplinary clusters. The tubes layout represents the transformation of cluster of discipline over time [28 (link)-30 ]. The width of tubes represents the number of records in which they appear in the same cluster. Darker tubes mean more disciplines are shared between 2 consecutive time periods.
Finally, 3 stages were completed, as follows, regarding the analysis of research hotspots. First, BICOMS was employed to calculate the frequency of keywords. Subsequently, a total of 13,706 keywords were obtained and merged based on the following 4 criteria [31 (link)]: (1) merging some keywords into corresponding Medical Subject Headings terms using PubMed (eg, “gynaecology” and “lymphadenectomy” were merged into “gynecology” and “lymph node excision,” respectively); (2) unifying the uppercase and lowercase of some keywords (eg, “Laparoscopy” and “Bladder cancer” were changed to “laparoscopy” and “bladder cancer,” respectively); (3) standardizing the singular and plural of keywords (eg, “child” and “pediatric” were changed to “children” and “pediatrics,” respectively); and (4) merging some synonym keywords (eg, “minimal invasive surgery” and “MIS” were replaced by “minimally invasive surgery”). After merging, 90 keywords with frequencies not less than 40 were obtained.
Second, we used BICOMS to generate the 88×88 co-occurrence matrix of keywords with a frequency not less than 40. It is worth noting that we removed robotic surgery and surgical robot because they are our research object. Then, a social network map was drawn with respect to these 88 keywords by Ucinet6.6 and VOSviewer [26 (link),32 ,33 (link)], which intuitively reflects the relationship between keywords of high frequency. The relative size of nodes is proportional to the frequency of keywords, whereas the relative width of lines is proportional to the correlation between keywords [34 (link)].
Third, we detected the burst strength of the cleaned keywords and drew a temporal bar graph for high-burst strength keywords. Burst strength depicts the intensity of the burst, that is, how great the change is in the word frequency that triggered the burst. Kleinberg burst detection algorithm [35 (link)] can recognize the sudden increase of word frequency over time and detect the burst of keyword popularity effectively. We chose Science of Science (Sci2) [36 ], which can implement this algorithm to find out the burst terms in the processed data and calculate the burst strength. Finally, 48 keywords with a burst strength of not less than 4 were obtained. However, these keywords may only be core keywords to a certain extent. Further screening by word frequency can improve the quality of core keywords. The higher the number of keyword frequency, the more likely it is to become a hot topic in future. Then we drew a temporal visualization map of 26 keywords with a frequency no less than 40 and burst strength more than 4 by Sci2 [37 (link)]. Each keyword has its own starting and ending time, and the area of each bar reflects its burst strength.