OPT is a three dimensional (3D) imaging technique that can produce 3D reconstructions of
transparent samples, requiring only a widefield imaging system and sample rotation. OPT can
be readily applied to chemically cleared samples, or to live transparent organisms such as nematodes
or zebrafish. For preclinical imaging, there is a trade-off between optical accessibility and
biological relevance to humans. Adult Danio rerio (zebrafish) represent a promising compromise,
with greater homology to humans than smaller animals, and superior optical accessibility
than mice. However, their size and physiology present a number of imaging challenges including
non-negligible absorption and optical scattering, and limited time for image data acquisition if
the fish are to be recovered for longitudinal studies. A key goal of this PhD thesis research was
to develop OPT to address these challenges and improve in vivo imaging capabilities for this
model organism.
This thesis builds on previous work at Imperial where angularly multiplexed OPT using
compressed sensing was developed and applied to in vivo imaging of a cancer-burdened adult
zebrafish, with a sufficiently short OPT data acquisition time to allow recovery of the fish after
anaesthesia. The previous cross-sectional study of this work was extended to a longitudinal
study of cancer progression that I undertook. The volume and quality of data acquired in
the longitudinal study presented a number of data processing challenges, which I addressed
with improved automation of the data processing pipeline and with the demonstration that
convolutional neural networks (CNN) could replace the iterative compressed sensing algorithm
previously used to suppress artifacts when reconstructing undersampled OPT data sets.
To address the issue of high attenuation through the centre of an adult zebrafish, I developed
conformal-high-dynamic-range (C-HDR) OPT and demonstrated that it could provide sufficient
dynamic range for brightfield imaging of such optically thick samples, noting that transmitted
light images can provide anatomical context for fluorescence image data.
To reduce the impact of optical scattering in OPT, I developed a parallelised quasi-confocal
version of OPT called slice-illuminated OPT (slice-OPT) to reject scattered photons and demonstrated
this with live zebrafish. To enable 3D imaging with short wave infrared (SWIR) light,
without the requirement of an expensive Ge or InGaAs camera, I implemented a single pixel
camera and demonstrated single-pixel OPT (SP-OPT) for the first time.