A Machine Learning-based Study Of Large-scale Glioma Literature Topics

Introduction:The body of glioma-related literature has grown significantly over the past 25 years.Despite this growth in the amount of published research,gliomas remain one of the most intransigent cancers.The purpose of this study was to analyze the landscape of glioma-related research over the past 25 years using machine learning and text analysis.Methods:In April 2019,we downloaded glioma-related publications indexed in PubMed between 1994 and 2018.We extracted the title,publication date,MeSH terms,and abstract from the metadata of each publication for bibliometric assessment.Latent Dirichlet allocation(LDA)was applied to the abstracts to identify publications' research topics with greater specificity.Results:We identified and analyzed a total of 52,625 publications in our study.We found that research on prognosis and the treatment of glioblastoma increased the most in terms of volume and rate of publications over the past 25 years.However,publications regarding clinical trials accounted for<5%.of all publications considered in this study.The current research landscape covers clinical,pre-clinical,biological,and technical aspects of glioblastoma;at present,researchers appear to be less concerned with glioblastoma' s psychological effects or patients' end-of-life care.Conclusion:Publication of glioma-related research has expanded rapidly over the past 25 years.Common topics include the disease's molecular background,patients'survival,and treatment outcomes;more research needs to be done on the psychological aspects of glioblastoma and end-of-life care.