Fig. 1.

VCID bibliometric topic map. Research topics are based on the results of the Latent Dirichlet Analysis (LDA) algorithm. LDA works by first creating a set of term vocabularies for each of a prespecified number of topics based on the terms’ co-occurrence in publication abstracts. The algorithm then uses those vocabularies to assign individual publications to one or more of these topics based on the frequency with which terms from that topic appear in the publication's abstract. Publications were pulled from PubMed using the following search string: (((dementia OR (cognitive impairment) OR (cognitive dysfunction)) AND (vascular OR cardiovascular OR cerebrovascular OR lacunar OR stroke)) OR CADASIL OR (cerebral autosomal dominant arteriopathy) OR (Binswanger[tiab] OR Binswanger's[tiab]) OR (cerebral amyloid angiopathy)) AND 2002:2016[dp]. For this analysis, the number of topics was set to 50, the algorithm was run on the publication abstracts in this data set. A topic similarity network was generated in which topics are connected if more than 45 articles in the data set were assigned to both of the connected topics. Descriptions were assigned to each topic based on the abstract and MeSH terms that most frequently appeared in each topic and on manual inspection of the papers assigned to each topic.