In pursuit of high‐fidelity waveguide imaging restoration using deep learning algorithms: A review

Abstract

AbstractWaveguide imaging is considered as one of the most important and widely used techniques in biomedical endoscopic applications. Recently, many attempts have been made to develop ever miniaturised in vivo imaging devices for minimally invasive clinical inspections. However, miniaturisation implies using a smaller optical aperture waveguide, which may introduce pixilation artefacts and pixel‐to‐pixel distortion to deteriorate overall imaging quality. To overcome the constraints imposed by miniaturised waveguides, the deep learning algorithms can be an effective tool to cure the imaging distortion via post‐processing, which already had encouraging results in many scenes of automatic machine‐learnt imaging restoration. The authors introduce the waveguide imaging transmission and the restoration algorithms, and then discuss their possible combinations. The results show that the integration of advanced waveguides and optimised algorithms can achieve unprecedented imaging restoration than before. In the future, in order to fill the need for high‐quality reconstructed images, we should not only improve ability of software to optimise restoration algorithms but also correspondingly concern hardware progress in waveguides. The practical sense of it is to help researchers better master and take advantage of these combinations to make next generation high‐fidelity endoscopes.