P120. A new computed assisted algorithm allows automatic evaluation of inflammatory activity in inflammatory bowel disease (IBD) using confocal laser endomicroscopy (CLE)
H. Neumann1, P. Mewes2, S. Foertsch3, D. Neumann2, M. Waldner1, M. Vieth4, M. Neurath1
1University of Erlangen-Nuremberg, Department of Medicine 1, Erlangen, Germany; 2Friedrich-Alexander University, Pattern Recognition Lab, Erlangen, Germany; 3University of Erlangen-Nuremberg, Department of Chemistry and Pharmacy, Erlangen, Germany; 4Klinikum Bayreuth, Institute of Pathology, Bayreuth, Germany
Background: Confocal diagnosis of acute inflammation in IBD is based on goblet cell depletion, vessel structures and crypt architecture. Nevertheless, interobserver agreement and intraobserver agreement in the assessment of these features is only moderate making a comparison of different studies nearly impossible. Aim of the study was to develop a new algorithm for the computed assisted diagnosis (CAD) of inflammatory activity in IBD in order to improve diagnostic reliability and comparability of CLE.
Methods: The CAD-algorithm semi-automatically segmented anatomical structures, like goblet cells, vessel structures, crypt architecture, and relevant cell types in CLE-image sequences. Therefore, the CAD-algorithm segmented and counted intestinal crypts and crypt lumens within each image. From the residual part of the image, image artifacts were automatically identified and afterwards excluded from further analysis using digital postprocessing. Subsequently, additional image filtering, segmentation and thresholding methods were applied to classify the different cell types and to measure a quantitative ratio.
Results: A new CAD-algorithm for the automatic evaluation of inflammatory activity in IBD using CLE was developed and evaluated. The CAD-algorithm allowed successful, fast and repeatable analysis of different architectural and cellular features, including goblet cells, crypt architecture and microvasculature with excellent overall accuracy.
Conclusions: The newly developed CAD-algorithm allowed automatic evaluation of inflammatory activity in IBD using CLE sequences. Thus, the technique has the potential to better characterize patients with IBD and to contribute to consistent data in different studies assessing inflammatory activity in IBD.