A comprehensive overview of acupuncture therapy over the past 20 years: Machine learning-based bibliometric analysis.

Background: Acupuncture, a therapy created by the ancient Chinese, has been gaining increasing popularity and acceptance worldwid. The surge in the number of publications on acupuncture therapy has posed significant challenges for researchers in effectively managing the vast amount of information. This study aimed to analyze the themes and topics of the scientific publications related to acupuncture therapy in the past two decades by machine learning.

Methods: The MeSH term "Acupuncture Therapy" was used for searching publications from 1st January, 2004-31 st December, 2023 on the PubMed database, while the R platform was adopted to obtain associated data. A topic network was constructed by latent Dirichlet allocation (LDA) and the Louvain algorithm.

Results: A total of 17,584 publications were finally recruited in this article. The publications were derived from 57 countries, with China, The United States and England being the top three countries. "Acupuncture Points", "Treatment outcome", "Electroacupuncture" were the most concerned MeSH terms. Four clusters and 50 branched topics were recognized by LDA and network analyses. "Comparative Efficacy", "Biochemical Analysis", "Symptomatic Treatment", "Professional Practice" and "Clinical Trials" are hotspots identified by LDA. "Neurotransmitter", "Meta-Analysis" and "Literature Review" have presented as new research hotspots.

Conclusions: Acupuncture therapy has obtained increasing attention over the past two decades. Most of the studies focus on the mechanisms especially the analgesic and anti-inflammatory mechanisms, more researches such as "Neurotransmitter" will continue to advance. Besides, "meta-analysis" and "literature reviews" are increasingly common, providing more comprehensive and credible evidence for acupuncture therapy.