A review of medical image segmentation research based on bibliometrics
Ruiting Li, Shuli ChengBackground
Medical image segmentation is a fundamental step in computer-aided diagnosis and treatment planning, and its research landscape has expanded rapidly with the advancement of artificial intelligence. A systematic bibliometric overview of this field is therefore needed to clarify its intellectual structure and emerging trends.
Objective
To provide a comprehensive bibliometric overview of the global research landscape, developmental trends, collaboration networks, and emerging hotspots in the field of medical image segmentation over the past decade.
Methods
A bibliometric analysis was performed on 21,437 publications related to medical image segmentation, indexed in the Web of Science Core Collection from 2016 to 2025. The analysis systematically examined publication outputs, country-level contributions, author collaboration networks, journal distributions, and keyword co-occurrence patterns.
Results
The annual publication count increased remarkably from 25 in 2016 to 5,207 in 2025. China contributed the largest share with 11,556 publications, followed by the United States (3,272) and India (1,924). Core researchers formed stable intra-team collaborations, though inter-team cooperation remained limited. Keyword and citation analyses identified deep learning, convolutional neural networks (CNNs), Transformers, U-Net, and their variants as dominant methodological themes. Case studies on skin and lung segmentation revealed differential research attention and international collaboration patterns across anatomical targets. A historical trajectory analysis showed methodological evolution from traditional image processing to fully convolutional networks and U-Net, and subsequently to hybrid architectures integrating attention mechanisms and Transformers.
Conclusion
Medical image segmentation has experienced substantial growth, driven by deep learning innovations. Future research may increasingly focus on multimodal foundation models capable of generalizable segmentation across diverse clinical scenarios. This bibliometric review offers researchers a valuable reference for understanding the intellectual structure and frontiers of the field.