Computational Optical Sectioning in Fibre Bundle Endomicroscopy

The field of fibre bundle endomicroscopy has emerged to enable real-time imaging of cellular level features in-vivo. The gold standard is confocal laserscanning, enabling optical sectioning. Point-scan confocal suffers from lower speeds, a need for complex alignment, and the added cost of a laser. This thesis presents three developments in computational optical sectioning for fibre bundle endomicroscopy.The first development is in structured illumination (SIM) endomicroscopy. Lower-cost, simplified endomicroscopes have been developed which use widefield incoherent illumination. Optical sectioning can be introduced to these systems using SIM. SIM improves imaging using spatial modulation of the focal plane and capturing a three-frame sequence. The acquired images are then numerically processed to reject out-of-focus light. This thesis reports and characterises the first high-speed SIM endomicroscope built using a miniature array ofmirrors, a digital micromirror device. The second development is automated motion compensation in SIM endomicroscopy. As a multi frame process, SIM is susceptible to motion artefacts, making the technique difficult to use in vivo and preventing the use of mosaicking to synthesise a larger effective field of view. I report and validate an automatic motion compensation technique to overcome motion artefacts and report the firstmosaics in SIM endomicroscopy.The third development is improvements in subtraction-based enhanced line scanning (ELS) endomicroscopy. The 2D scanning of a point scan confocal endomicroscope can be replaced by a scanning line which is synchronised to the sequential readout of a rolling shutter camera. While this leads to high-speed sectioning, as with all line scanning systems, far-from-focus light degrades images. It is possible to remove this by subtracting a second image taken with an offset detection slit. This has previously required two-cameras or two sequentialframes. The latter introduces motion artefacts. This thesis presents a novel approach to ELS using single frame acquisition with real-time mosaicking at 240frames/s.