Abstract |
The use and development of spectroscopic techniques with emphasis on advanced microscopic methods for
the timely and accurate diagnosis of diseases such as cancer or the need to understand some of the most
important mechanisms of Biology, such as aging, based on observation and extraction of information at the
cellular level is an interdisciplinary field of enormous research interest. Non-linear imaging microscopy is
a reliable diagnostic tool for microscopic studies, as it is a minimally invasive technique with high
resolution, while also providing the ability to simultaneously capture images through different non-linear
signals such as: second and third harmonic generation (SHG / THG) or multiphoton excitation fluorescence
(MPEF). During the present dissertation, studies were carried out that help the further development of the
above techniques for the investigation of various biological problems. In addition, the creation of new data
analysis algorithms and the implementation of artificial neural networks open new horizons in the fields of
application of nonlinear microscopy.
Specifically, the first of the above studies concerns the application of neural networks to THG images from
human biopsy specimens for the accurate diagnosis of breast cancer at subcellular level. Although only the
diagnosis of this type of cancer based on the imaging of unstained biopsies through THG was very
innovative, it faced a key problem, the need for specialized staff to evaluate and separate the samples. THG
mainly depicts optical inhomogeneities, lipids or cell membranes and the diagnosis is based on various
characteristics of the cells such as their volume, their morphology and their emitted signal. However,
locating cells in THG images is an extremely time consuming process and requires special experience, thus
greatly limiting the use of this non-invasive diagnostic technique. The solution to this problem was given
through the application of neural networks to categorize specific images. It is worth noting that the
application of neural networks in THG images for the diagnosis of breast cancer was done for the first time
and its success leads to the elimination of the need for specialized staff and the dramatic reduction of
diagnostic time, thus bringing non-linear microscopy one step closer to clinical trials.
The second study also concerns the diagnosis of breast cancer based on the above biopsies and non-linear
microscopy, but this time the recorded non-linear signal was that of SHG. Tissues are rich in collagen fibers
which is capable of emitting strong SHG signals. In recent years there have been several studies on the
dependence of SHG on incoming polarization of radiation (PSHG) and on the quantified information
obtained through it, such as the anisotropy parameter B. Through this parameter but also with the
introduction of a new one (ratio parameter) it became possible to completely differentiate between all cancer
stages of the biopsies we studied (from benign to third stage). The calculation of both of these parameters
takes less than one second for each sample as it is based on the Fourier analysis of PSHG measurements.
In addition, a biophysical model was proposed that interprets these results based on the mechanical strain
applied to the collagen fibers during the various stages of cancer.
In this dissertation, the capabilities of PSHG in combination with Fourier analysis were also used to study
possible structural changes in the striated muscles of the model organism Caenorhabditis elegans (C.
elegans) during aging. The results of this study indicate that the striated muscle structure of C. elegans
changes as the age of the sample increases. As far as is known, this is the first time that differentiation of
the PSHG results has been observed through Fourier analysis from in-vivo muscle as the sample age
increases. The differentiation was based on the difference of the spectral phases of the recorded PSHG
signals. In addition, through the newly introduced ratio parameter, this study showed that the hitherto usual
assumption of cylindrical symmetry for the biophysical model of PSHG lacks in satisfactory and complete
description of the recorded data compared to the triangular symmetry, which in turn is inferior to the most
general case where the sample does not need to have any particular symmetry.
The last part of this thesis was dedicated to highlight the prospect of non-linear microscopy for application
in material studies with the aim of delivering innovative results. Νon-linear measurements were used to
study Cultural Heritage art works. Specifically, a specially designed algorithm was developed, which
through MPEF measurements determines with great accuracy the thickness of the protective varnish layer
in a work of art, thus helping in the process of its restoration.
|