Commentary: A quantitative analysis of artificial intelligence research in cervical cancer: a bibliometric approach utilizing CiteSpace and VOSviewer

Abstract

Introduction In recent years, the exponential growth in biomedical literature has garnered significant attention for bibliometrics as a method capable of quantitatively and qualitatively analyzing research trends and hotspots within a given discipline. We read with great interest on the publication by Ziqi Zhao et al. (1) titled "A quantitative analysis of artificial intelligence research in cervical cancer: a bibliometric approach utilizing CiteSpace and VOSviewer", which has published in the issue of Frontiers in Oncology.We highly support and appreciate the researchers' work, and thank them for their contributions in the field. Commentary and discussion By using bibliometrics, this study conducted an in-depth analysis of existing publications on the application of artificial intelligence (AI) in the field of cervical cancer. The analysis reveals that AI technology is playing an increasingly important role in several key aspects of cervical cancer, including early diagnosis screening, treatment plan formulation, prognostic evaluation, and image analysis. However, we identified several points requiring clarification and correction. First, inconsistencies in Literature Search and Screening Numbers: the manuscript states in multiple sections that 927 publications were ultimately included. However, Figure 1 (Flow chart of the retrieval process) indicates that:97 publications were excluded for not being original research or review articles and 6 publications were excluded for being non-English. This results in a total exclusion of 103 publications. Therefore, the initial number of records identified must logically be 927 (included) + 103 (excluded) = 1030 records. This calculation conflicts with Table 1 which states "research results from SSCI and SCI-E (N=1032) and the "Manual screening process "section which states "we preliminarily screened 1,027 relevant papers". We recommend the authors verify and correct these inconsistencies (1032 in Table 1 and 1027 in the text) to align with the flow chart data, which implies an initial count of 1030 records. Second, omission in Institution Ranking (Table 2):The "Productive institutions analysis" section states: "According to the data in Table 2, the top three institutions in terms of TLS are the National Cancer Institute of the National Institutes of Health (TLS=141), Southern Medical University (TLS=81), and the Chinese Academy of Medical Sciences (TLS=80)."However, examination of Table 2 reveals that Peking Union Medical College also has a TLS=80, placing it equally third with the Chinese Academy of Medical Sciences. We recommend the authors amend this sentence to acknowledge Peking Union Medical College as co-third place. Third, concerns Regarding CiteSpace g-index Parameter (k-value):The methodology for generating the institutional collaboration network (Figure 4B) states: "...setting a time span from 2008 to 2024, with a 1-year slicing length, using institutions as the node type, and setting the g-index to k=8".We note that Figure 3D uses a g-index of k=25, while all other figures generated with CiteSpace in this study reportedly use k=10.It is understood that in CiteSpace, the g-index parameter (k) controls the number of nodes selected within each time slice. A higher k-value (k=25) retains more nodes, while a lower k-value (k=8 or k=10) retains fewer nodes. We are concerned that the use of k=8 (Figure 4B) and k=10 (most other figures) may be significantly lower than the common default or standard value (often k=25), potentially excluding too many nodes. This raises the question: Could these relatively low k-values have resulted in an incomplete representation of the networks, failing to fully capture and interpret the relevant collaborative structures or knowledge domains? We recommend the authors justify their choice of these specific k-values and discuss whether this parameter selection might have impacted the comprehensiveness of their network visualizations and analyses.

Ähnliche Arbeiten