Bibliometric Analysis

Bibliometrics is the quantitative study of literature and a measurable method used to identify the developmental trends within a certain field to obtain quantifiable, reproducible, and objective data. In this analysis, we computed the growth rate of publications, characteristics of research activities (topics and keywords), publication patterns (countries and journals), and research hotspot tendencies (citation bursts and timeline map).

The growth rate of publications over time was computed by raising the rate of the number of publications in 2019 over the number of publications in 1995 to the power of 1/24, as shown below. The publication trends of the number of publications each year were also reported.

Growth rate = ((number of publications in the last year ÷ number of publications in the first year)^{1/(last year - first year)} - 1) × 100

Citation trends of the top 10 countries, top 10 journals, and top 10 research domains were analyzed for the publication patterns. Frequencies and percentages of publications in each journal and each country were computed based on the publication years. The countries were identified by the affiliations of the listed authors. This information was provided by WoS. The rank of research impact of each country and each journal was provided based on the citation rates.

The characteristics of research activities were analyzed according to the topics and the keywords of publications. Top topics, including health problems, AI technology, function, and population, were identified and described by frequency, percentage, and citation rate of keywords listed by the authors. The centrality of a keyword was a combination of statistic equations that measure the representativeness of selected words for the text content according to betweenness, closeness, degree, eigenvector, PageRank (Google LLC), eccentricity, coreness, clustering coefficient, and term frequency scores [16]. The centralities of keywords were computed using HistCite. .

Citation bursts and a timeline map were developed using the HistCite software. The cluster view was generated based on publications between 1995 and 2019, and each cluster was labeled by the keywords used by the paper. The time slice was set as 1 year and the threshold interpolation citations, cocitations, and cocitation coefficient were set as 4, 1, and 20, respectively. A minimum tree calculation formula was adopted to tailor the network.