Minimally invasive computational 3D lensless fiber endomicroscopy

Abstract

Endoscopy is essential for biomedical imaging and diagnosis as well 3D Metrology and process monitoring in confined spaces. Conventionally 3D endoscopes are limited in the minimally achievable diameter by distal scanning or stereo optics to several millimetres. Recent progress and perspectives towards 3D submillimetre endoscopes using coherent fibre bundles (CFBs) is discussed. While CFBs allow for diameters of only a few 100 µm, they are commonly used for relaying intensity patterns, only. This is due to the complex-valued optical transfer function (OTF), which results in a time and wavelength depended phase scrambling. Furthermore, CFBs offer only a few 10,000 fibre cores limiting the achievable space-bandwidth product. Different approaches based on holography, digital optical phase conjugation, 2P-Polymerisation-based 3D printing and deep learning for enabling lensless 3D single-shot imaging with sub-micron resolution are introduced and compared. Optical OTF compensation can be realized using diffractive optical elements (DOEs). For static distortions, static DOEs printed by 2P-polymerization are sufficient. While dynamic distortion require an in-situ calibration and programmable DOEs such as spatial light modulators. OTF independent compensation can be realized by a low coherence common path interferometry. A second approach is based on circumventing the complex valued OTF by evaluating intensity information only, just like conventional lens based endoscopes. In order to achieve capability, an optical diffuser substitutes the distal lens. The diffuse scattering of light is used to code 3D object information in speckle patterns and relay this information through the CFB. Deconvolution algorithms or artificial neural networks can perform the decoding on the distal side numerically. The novel application areas for these single shot 3D measurements will be highlighted.