OCT A-Scan Based Lung Tumor Tissue Classification With Bidirectional Long Short Term Memory Networks
Sebastian Otte, Christoph Otte, Alexander Schlaefer, Lukas Wittig, Gereon Hüttmann, Daniel Drömann, Andreas Zell

Abstract:
This paper presents a novel method for lung tumor tissue classification using Bidirectional Long Short Term Memory networks (BLSTMs). Samples are obtained through Optical Coherence Tomography (OCT) from real soft tissue specimen and represented as data sequences. Such sequences are learned with BLSTMs, which are able to recognize even non-uniformly compressed temporal encoded patterns in sequential data in both forward and backward time-direction. Our experiments indicate that BLSTMs are a suitable choice for this classification task, since they outperform other recurrent architectures. Furthermore, the presented findings lead to promising future investigations in the field of OCT based tissue analysis.