PT Chapter
AU Michal Drozdzal
   Santiago Segui
   Petia Radeva
   Carolina Malagelada
   Fernando Azpiroz
   Jordi Vitria
TI An Application for Efficient Error-Free Labeling of Medical Images
BT Multimodal Interaction in Image and Video Applications
PY 2013
BP 1
EP 16
VL 48
DI 10.1007/978-3-642-35932-3_1
AB In this chapter we describe an application for efficient error-free labeling of medical images. In this scenario, the compilation of a complete training set for building a realistic model of a given class of samples is not an easy task, making the process tedious and time consuming. For this reason, there is a need for interactive labeling applications that minimize the effort of the user while providing error-free labeling. We propose a new algorithm that is based on data similarity in feature space. This method actively explores data in order to find the best label-aligned clustering and exploits it to reduce the labeler effort, that is measured by the number of “clicks. Moreover, error-free labeling is guaranteed by the fact that all data and their labels proposals are visually revised by en expert.
ER