Synthetic Data and Generative AI in Oncology: A Global Bibliometric and Science-Mapping Analysis of 5,687 Publications
Mohammad Ghosheh, Mawada Hussein, Balqees Mansour, Dhiyaa Aljorf, Amin Abyad, Ghadeer GhoshehAbstract
Background:
The field of synthetic data has emerged as a transformative instrument in oncology, addressing the persistent constraints of patient-data scarcity, privacy regulation, and class imbalance that impede artificial intelligence development. Despite accelerating output, no comprehensive bibliometric analysis has mapped this field across the full oncological spectrum, with the sole prior study being confined to genomic cancer medicine.
Objective:
This study aims to characterize the intellectual structure, productivity, and conceptual evolution of synthetic data research in oncology.
Materials and Methods:
A total of 5687 articles (1972–2026) were retrieved from Scopus using a refined Boolean strategy. Performance analysis, collaboration mapping, and science mapping were conducted using the bibliometrix R package, VOSviewer, and CiteSpace, encompassing Bradford’s law, thematic evolution, conceptual mapping, co-occurrence networks, and citation-burst detection.
Results:
The corpus exhibited a 12.52% annual growth rate and logistic trajectory approaching maturity. China and the United States dominated production and citation impact, respectively, while the Chinese Academy of Sciences led the institutional output. Collaboration networks displayed a pronounced core–periphery structure. Bradford’s law confined one-third of the output to 21 core journals. Deep learning constituted the methodological core; thematic and trend analyses disclosed a consistent transition from statistical and genomic methods toward generative adversarial networks, diffusion models, and privacy-preserving paradigms, with breast cancer as the enduring clinical anchor.
Conclusion:
Synthetic data research in oncology has consolidated into a coherent, deep learning-driven field. Future work should extend synthetic methods beyond imaging, validate fidelity and privacy, and redress geographical asymmetries to advance precision oncology.