Aristotle University Of Thessaloniki (AUTH)
Υπολογιστικές μεθοδολογίες τυποποίησης ιών
Abstract
dc:descriptionThe continuous mutations that take place in the genetic material of a certain virus constantly trigger the creation of new types of the virus. Because of these mutations, the new virus types obtain features that differentiate the severity of their infection (e.g., the human papillomavirus which appears in both oncogenic and non-oncogenic types). In these cases the identification of the exact type(s) of a virus that have infected a patient, namely the task of typing the virus, has acquired great importance with regards to the efficient treatment of its infection. Due to the importance of virus typing, several molecular biology methods that aim at efficiently discriminating among the various types of a virus based on their genotypic differences have been developed during the last decades. Among them, the molecular method called PCR-RFLP gel electrophoresis is currently employed widely over the world. However, the procedure of virus typing via the discussed method remains, in contrast to the other methods, heavily manual. This shortcoming of the PCR-RFLP method with respect to automation makes the conventional typing protocol of the method error-prone – especially in complex cases of multiple infections – and laborious for the molecular biologists who execute it. The aforementioned issues of the PCR-RFLP method can be tackled with the help of digital signal processing techniques along with the estimation and decision theory. For this purpose, the present thesis develops a series of novel computational methodologies that undertake the entire task of virus typing via the PCR-RFLP method in a consistent and effective manner. In order to ensure the correctness of the typing decisions, the proposed methodologies employ additional information that has been ignored by the conventional typing protocol, namely the concentration of the viral genetic material. The introduction of a factual observation model of the PCR-RFLP examination product sets the ground for the proposed methodologies to employ the additional information. On top of this model, the present thesis develops an integrated computational typing methodology (the static methodology), which analyzes the examination product and reaches typing decisions. The performance of the static methodology is evaluated thoroughly through theoretic investigations and experiments on real as well as simulated typing data. Moreover, the static methodology is implemented in its entirety as a graphical software application. Apart from the static methodology, the present thesis introduces also the dynamic typing methodology, which is capable of making typing decisions while the PCR-RFLP examination is still in progress. Dynamic typing is achieved through the establishment of communication between the previously isolated phases of the conventional typing protocol and it tackles several open issues associated with the static approach. An integrated hardware-software system is designed in order to implement the dynamic methodology. The feasibility of the dynamic typing approach is validated through an experiment that emulates the operation of the aforementioned system. Both the proposed typing methodologies are complemented by a novel algorithm (the fast algorithm) which undertakes the most demanding part of the typing procedure in terms of computational complexity, namely the task of testing the possible combinations of virus types. The fast algorithm parcels the aforementioned task into a set of subproblems; this way, the typing procedure is significantly accelerated without essentially harassing the typing accuracy. The fast algorithm is evaluated extensively on a well-sized set of real typing data. Before exploiting the concentration information as suggested by the proposed typing methodologies, the background intensity of the images that result from the PCR-RFLP examination needs to be efficiently removed. For this purpose, a novel method that models the background intensity of an examination image as a polynomial of the space coordinates and then successfully removes it from the image is introduced in the present thesis.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2011
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Μαραμής, Χρήστος
- Maramis, Christos
Subjects
dc:subject × 18- Βιοϊατρκή μηχανική
- Ψηφιακή επεξεργασία σήματος
- Θεωρία εκτίμησης & απόφασης
- Τυποποίηση ιών
- Υπολογιστικά υποβοηθούμενη διάγνωση
- Ολοκληρωμένη μεθοδολογία
- Μέθοδος PCR-RFLP
- Biomedical engineering
- Digital signal processing
- Estimation & decision theory
- Virus typing
- Computer-aided diagnosis
- Integrated methodology
- PCR-RFLP method
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Natural Sciences
- Computer and Information Sciences
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/28494