On-the-Fly Detection of Changes in Mucosal Tissue Architecture During Endoscopy - A Flag as to Where and When to Take Biopsies

In this paper we present a segmentation technique that raises a flag on-the-fly as to when a transition occurs between different mucosal architectures while traveling along the surface of the tissue for careful and subsequent examination. The presented segmentation technique has the potential to enhance the endoscopist's ability to locate and identify abnormal mucosal architectures; and to help in the decision as to when and where to take biopsies; steps that should lead to improvement in the diagnostic yield. The segmentation scheme is based on detecting an abrupt change in the parameters extracted from the stochastic decomposition method (SDM) that models the scattered signal. Subsequent evaluation of the homogeneous regions separated by the found edges for further mucosal type classification is then possible. The tests are performed on two types of animal tissue data collected from rat colon and rabbit colon in vitro. The results demonstrate the effectiveness of the proposed segmentation scheme to detect the transition between different mucosal structures on-the-fly with sensitivity reaching 40 mum which corresponds to the sliding segmentation window step size used in our experiments.