Fibre-based inelastic microspectroscopy - towards Brillouin endoscopy
File(s)
Author(s)
Xiang, Yuchen
Type
Thesis
Abstract
A major incentive of modern bio-imaging is to make use of optically sensitive properties to accurately diagnose diseases that are usually life threatening. A key technological advancement in recent years is the emergence of hyperspectral imaging, which intelligently combines spectroscopy with conventional imaging. In turn, information from the images is diversified and enhanced, which is enabled by more biologically sensitive and specific contrast mechanisms. Amongst the newly available avenues of contrast, imaging with Brillouin spectroscopy, is uniquely promising for its ability to yield mechanically-specific information in a non-contact manner. While the possibilities are numerous, this work considers specifically the feasibility of constructing an endoscopic probe for \emph{in vivo} detection of cardiovascular diseases using fibre-based Brillouin instrumentation.
After summarising the main design challenges, it was deduced that the main obstacle was the fibre-induced background. Two proof-of-concept designs were thus presented experimentally, which explored options with and without additional filtering, preliminary results were obtained from common liquids in both cases. The first kind of design was preferable for its intrinsically high collection efficiency. The requirement of extensive filtering, however, was undesirable and the device was ultimately observed to be severely limited by the SNR, due to a combination of residual background noise and lossy fibre optical components. The second type of design did not require additional filtering in contrast and registered good signal levels in a range of liquids remotely. This, however, was only facilitated by the increase in acquisition time, which was deemed infeasible for \emph{in vivo} applications and no further improvement was viable due to the intrinsically lossy off-axis collection geometry. While novel optical solutions exist, which demand specialised facilities, the attention was shifted towards software-based techniques for SNR optimisation. Namely, two reconstruction algorithms - using maximum entropy and wavelet analysis, were investigated and demonstrated good ability in denoising the spectra. This enabled the superior estimation of more accurate Brillouin parameters through noise. Finally, the outlook of \emph{in vivo} HS imaging was explored with the implementation of various multivariate algorithms to form images of Adipocyte cells, the results prove these methods to be faster, more accurate and more noise-resilient.
After summarising the main design challenges, it was deduced that the main obstacle was the fibre-induced background. Two proof-of-concept designs were thus presented experimentally, which explored options with and without additional filtering, preliminary results were obtained from common liquids in both cases. The first kind of design was preferable for its intrinsically high collection efficiency. The requirement of extensive filtering, however, was undesirable and the device was ultimately observed to be severely limited by the SNR, due to a combination of residual background noise and lossy fibre optical components. The second type of design did not require additional filtering in contrast and registered good signal levels in a range of liquids remotely. This, however, was only facilitated by the increase in acquisition time, which was deemed infeasible for \emph{in vivo} applications and no further improvement was viable due to the intrinsically lossy off-axis collection geometry. While novel optical solutions exist, which demand specialised facilities, the attention was shifted towards software-based techniques for SNR optimisation. Namely, two reconstruction algorithms - using maximum entropy and wavelet analysis, were investigated and demonstrated good ability in denoising the spectra. This enabled the superior estimation of more accurate Brillouin parameters through noise. Finally, the outlook of \emph{in vivo} HS imaging was explored with the implementation of various multivariate algorithms to form images of Adipocyte cells, the results prove these methods to be faster, more accurate and more noise-resilient.
Version
Open Access
Date Issued
2020-06
Date Awarded
2021-01
Copyright Statement
Creative Commons Attribution-Non Commercial 4.0 International Licence
License URL
Advisor
Török, Peter
Paterson, Carl
Sponsor
Engineering and Physical Sciences Research Council
Publisher Department
Department of Physics
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)