Building a mass online annotation tool for dental radiographic imagery

Thomas Ekert, Joachim Krois, Falk Schwendicke

Abstract: In medical images analysis in general and for dental radiographic imagery more specifically, deep learning algorithms, in particular convolutional neural networks, are increasingly applied for detecting pathologies. However, creating an sufficiently large data set of labeled radiographic imagery for training deep neural networks is resource- and time-consuming. In order to build a large and reliably database of annotated dental radiographic imagery we are building a mass online annotation tool (Mono) for dental panoramic x-ray images. The tool is designed to allow evaluating the annotator’s proficiency against clinical records, majority vote schemes and model predictions in real time.

Author Affiliation: CODE University of Applied Science and Charité – Universitätsmedizin Berlin

Keywords: online annotation system, radiographic imagery, oral pathologies, dentistry