Overview of the CLEF 2010 medical imageretrieval track

Abstract

It was decided to hold the seventh round of the ImageCLEF medical retrieval assignment in 2010. Like the collections in 2008 and 2009, the 2010 collection includes photographs and captions from the Radiology and Radiographics journals published by the RSNA (Radiological Society of North America) as well as other sources (Radiological Society of North America). Three sub–tasks were completed under the auspices of the medical task: modality identification, image–based retrieval, and case–based retrieval, all of which were completed under the auspices of the medical task. The goal of the modality identification task was to ascertain the mode of acquisition of the images in the collection by utilising visual, textual, or a combination of approaches to identify them. When performing an image–based retrieval task, the goal was to return an ordered set of images from the collection that best met the information need specified as a textual statement and as a set of sample images, whereas when performing a case–based retrieval task, the goal was to return an ordered set of articles (rather than images) that best met
the information need specified as a description of a "case." The number of research organisations registering for the medical task has increased to 51 from the previous number of registrants. The number of groups submitting runs, on the other hand, has remained constant at 16, with the total number of submitted runs increasing to 155. Ad hoc runs made up 51 of them, while case–based runs made up 48 and modal–ity classification runs made up the remaining 46 (see table below). The best results for the ad–hoc retrieval themes were obtained via the use of mixed approaches, with textual techniques also providing satisfactory results. For the case–based topics, textual means were unquestionably preferable. While textual and visual tactics alone were somewhat successful in the modality de-tection test, it was the combination of these approaches that proved to be the most effective