Magnetically-guided capsule endoscopy (MGCE) was introduced in 2010 as a procedure where a capsule in the stomach is navigated via an external magnetic field. The quality of the examination depends on the operator's ability to detect aspects of interest in real time. We present a novel two step computer-assisted diagnostic-procedure (CADP) algorithm for indicating gastritis and gastrointestinal bleedings in the stomach during the examination. First, we identify and exclude subregions of bubbles which can interfere with further processing. Then we address the challenge of lesion localization in an environment with changing contrast and lighting conditions. After a contrast-normalized filtering, feature extraction is performed. The proposed algorithm was tested on 300 images of different patients with uniformly distributed occurrences of the target pathologies. We correctly segmented 84.72% of bubble areas. A mean detection rate of 86% for the target pathologies was achieved during a 5-fold leave-one-out cross-validation. © 2011 Springer-Verlag.
CITATION STYLE
Mewes, P. W., Neumann, D., Licegevic, O., Simon, J., Juloski, A. L., & Angelopoulou, E. (2011). Automatic region-of-interest segmentation and pathology detection in magnetically guided capsule endoscopy. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 6893 LNCS, pp. 141–148). https://doi.org/10.1007/978-3-642-23626-6_18
Mendeley helps you to discover research relevant for your work.