Landscape of surgery in Crohn's disease across twenty years: insights from machine learning.

Background: Crohn's disease continues to be a major component of inflammatory bowel disease with increasing incidence and prevalence. Increasing publications of surgery in Crohn's disease have significantly expanded the research scope. The aim of this study is to characterize main topics and a full landscape of surgery in Crohn's disease. Methods: Studies of surgery in Crohn's disease from 2000 to 2020 were screened and retrieved from the Web of Science Core Collection database. Latent Dirichlet allocation (LDA), one of machine-learning algorithms for natural language processing, was employed for topic modeling. All the studies were processed, analyzed and visualized by R software, CiteSpace and Gephi. Results: A total of 3,697 original publications were identified from the database. USA was the leading country with the most top institutions such as Cleveland Clin Florida and Mayo Clinic and Mayo Foundation. Increasing impact of institutions from Korea and China was also noticed. Bo Shen was the leading author in publication. A machine learning based topic modeling identified major clusters, including disease assessment, surgical treatment and complications, risk factors and epidemiology, disease development and diagnosis, target treatment and recurrence. Three topics attracted continuous high research attention, including expression of intestinal cell, perianal fistula and laparoscopic and open operation. Conclusions: This study identified key topics relating to the development of surgery in Crohn's disease, and provided bibliometric insights and perspectives for future development in the field of surgery in Crohn's disease.