Machine Learning-Based Text Processing Reveals Research Trends in Heterotopic Ossification.

BACKGROUND

Heterotopic ossification (HO) represents a highly active research field in pathological bone formation. Despite substantial advancements, a comprehensive understanding of its underlying mechanisms and clinical trajectory remains incomplete.

MATERIALS AND

METHODS

A bibliometric and machine-learning latent Dirichlet allocation (LDA) analysis was performed using data retrieved from the Web of Science Core Collection database.

RESULTS

A total of 4722 publications related to HO were identified. The most prolific countries included the USA, CHINA, GERMANY, the UK, JAPAN, SOUTH KOREA, ITALY, TURKEY, FRANCE, and CANADA . GERMANY demonstrated the highest citation strength, followed by CHINA . Aside from "heterotopic ossification," other frequently occurring author keywords included "fibrodysplasia ossificans progressive," "complications" and "total hip arthroplasty." In keywords plus, besides HO, replacement, bone-formation, and arthroplasty were the most frequently occurring terms. Institutional network analysis with subject-specific clustering indicated that Shanghai Jiao Tong University was significantly enriched in radiology, nuclear medicine and medical imaging, while Wilderness Spine Serv specialized in surgical management. A developmental timeline plot of a network of most contributing authors also was visualized, along with the most influential references. Meanwhile, citation analysis indicated that Kaplan FS and Shore EM were the top-cited authors. By LDA analysis, a total of 16 key topics were identified in this field with distinct period-proportion visualization. One of the topics, cell bone express differentiation and formation has clearly dominated the last 10 years.

CONCLUSION

This study constitutes the most extensive text processing analysis of HO to date, offering valuable insights and directions for future development.