Results - Details
Search command : Author="Οικονόμου"
And Author="Αθανάσιος"
Current Record: 1 of 4
|
|
| Title |
Methods for melanin quantification on fish scales and color regression on Porgus skin using Hybrid Microscopy and Machine Learning |
|
Author
|
Οικονόμου, ΑΘανάσιος
|
|
Thesis advisor
|
Τσιρώνης, Γεώργιος
|
|
Reviewer
|
Ζαχαράκης, Γιάννης
Παυλίδης, Μιχαήλ
|
| Abstract |
Fishes display sophisticated skin chromatic properties that are of considerable ecological,
physiological and behavioural importance. From behavioral regulations, to crypsis and
adaptiveness to light, color patterns play an important role to the wellbeing of fishes.
Contemporary methods of studying this subject either at a microscopic or macroscopic level, have
many disadvantages and drawbacks. In this work, combining three disciplines, Physics, Computer
Science and Biology, we propose novel procedures that measure and analyse the chromatic
properties of fish skin, either at a microscopic or macroscopic level. Using a hybrid photoacoustic
and fluorescence configuration we were able to create a method to quantify melanin concentration
on a fish scales. Hybrid microscopy’s advantages are clearly demonstrated in our results. The
complementarity of such a configuration creates datasets with information properly separated and
easy to work with. Furthermore, photoacoustic microscopy in particular, gives us quantified
information due to the fact that each signal on an image is proportional to the melanin
concentration. Using Machine Learning and image processing algorithms we were able to analyse
the quantified results and prove statistical important, intra class, differences. A Convolutional
Neural Network was created in order to automatically distinguish an input to its respective class,
opening a road to further investigations on the subject. Finally, on a macroscopic level with the
use of image processing and Convolutional Neural Networks, a model was created and trained
that was able to robustly calculate the chromatic parameters on specific positions on a fish's
body, possibly replacing the use of a colorimeter, the go to instrument when it comes to color
measurements. Every model and method that is demonstrated here, is fully automated hoping to
create useful and valuable tools for biologists studying the respective fields.
|
| Language |
English |
| Issue date |
2021-07-28 |
| Collection
|
School/Department--School of Sciences and Engineering--Department of Physics--Graduate theses
|
|
|
Type of Work--Graduate theses
|
| Permanent Link |
https://elocus.lib.uoc.gr//dlib/2/2/4/metadata-dlib-1625571472-690667-9151.tkl
|
| Views |
207 |
Search
