User-centered image indexing—often reported in research on collaborative tagging, social classification, folksonomy, or personal tagging—has received a considerable amount of attention [1-7]. The general themes in more recent studies on this topic include user-centered tagging behavior by types of images, pros and cons of user-created tags as compared to controlled index terms; assessment of the value added by user-generated tags, and comparison of automatic indexing versus human indexing in the context of web digital image collections such as Flickr. For instance, Golbeck's finding restates the importance of indexer experience, order, and type of images . Rorissa has found a significant difference in the number of terms assigned when using Flickr tags or index terms on the same image collection, which might suggest a difference in level of indexing by professional indexers and Flickr taggers . Studies focusing on users and their tagging experiences and user-generated tags suggest ideas to be implemented as part of a personalized, customizable tagging system. Additionally, Stvilia and her colleagues have found that tagger age and image familiarity are negatively related, while indexing and tagging experience were positively associated .
A major question for biomedical image indexing is whether the results of the aforementioned studies, all of which dealt with general image collections, are applicable to images in the medical domain. In spite of the importance of visual material in medical education and the prevalence of digitized images in formal medical practice and education, medical students have few opportunities to annotate biomedical images. End-user training could improve the quality of image indexing and so improve retrieval. In a pilot assessment of image indexing and retrieval quality by medical students, this study compared concept completion and retrieval effectiveness of indexing terms generated by medical students on thirty-nine histology images selected from the PubMed Central (PMC) database. Indexing instruction was only given to an intervention group to test its impact on the quality of end-user image indexing.
Digital Object Identifier (DOI)
Kim, Sujin, "An exploratory study of user-centered indexing of published biomedical images" (2013). Pathology and Laboratory Medicine Faculty Publications. 1.