Tissue segmentation in volumetric laser endomicroscopy data using FusionNet and a domain-specific loss function

摘要

Volumetric Laser Endomicroscopy (VLE) is a promising balloon-based imaging technique for detecting earlyneoplasia in Barretts Esophagus. Especially Computer Aided Detection (CAD) techniques show great promisecompared to medical doctors, who cannot reliably find disease patterns in the noisy VLE signal. However, anessential pre-processing step for the CAD system is tissue segmentation. At present, tissue is segmented manuallybut is not scalable for the entire VLE scan consisting of 1,200 frames of 4,096 on 2,048 pixels. Furthermore,the current CAD methods cannot use the VLE scans to their full potential, as only a small segment of theesophagus is selected for further processing, while an automated segmentation system results in significantly moreavailable data. This paper explores the possibility of automatically segmenting relevant tissue for VLE scansusing FusionNet and a domain-specific loss function. The contribution of this work is threefold. First, we proposea tissue segmentation algorithm for VLE scans. Second, we introduce a weighted ground truth that exploits thesignal-to-noise ratio characteristics of the VLE data. Third, we compare our algorithm segmentation againsttwo additional VLE experts. The results show that our algorithm annotations are indistinguishable from theexpert annotations and therefore the algorithm can be used as a preprocessing step for further classification ofthe tissue.