This paper addresses the issue of a content-based information retrieval system that works on fMRI images from neuroscientific journal publications. We present a general framework for automatic extraction, characterisation and classification of fMRI images, based on their functional properties. The proposed method identifies the section of each of those images, by morphological processing, and estimates the coordinates of the brain activated regions, in relation to a standard reference template using locality preserving projections. Those regions are then segmented, and their physical and geometrical properties evaluated. We formulate a feature vector based on these characteristics, and cluster the images and corresponding journal publications using self organizing maps. © 2010 Springer-Verlag Berlin Heidelberg.
CITATION STYLE
Rajasekharan, J., Scharfenberger, U., Gonçalves, N., & Vigário, R. (2010). Image approach towards document mining in neuroscientific publications. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 6065 LNCS, pp. 147–158). https://doi.org/10.1007/978-3-642-13062-5_15
Mendeley helps you to discover research relevant for your work.