Democritus University of Thrace (DUTH)
Ευφυείς μέθοδοι επεξεργασίας ψηφιακής εικόνας του βασικοκυτταρικού καρκινώματος βλεφάρου
Abstract
dc:descriptionThe Basal Cell Carcinoma (BCC) is known to be a slowly growing, locally expanding, malicious epidermal tumor (Telfer et al, 1999). It constitutes the most common epidermal cancer. It is mostly caused due to the exposition to the Sun. On the field of curing this kind of cancer, there are a lot of types of treating the tumor. It usually is necessary that the tumor to be removed surgically. Besides that, there are some treatments that could take place, like cryotherapy, photodynamic therapy, or even with medicines (Slade et al, 1998), but in order to choose the best method, according to the type of BCC, there are some tests to be done and some cosiderations regarding the age, clinical condition of the BCC and other characteristics of the patient, so that the right method is chosen.The first step of curing a tumor is to identify it in the early stages of its evolution. In this direction there is a lot of effort that has been done on developing an Intelligent System using Neural Networks, that could identify a BCC even in the early stages.This is the subject of this project. With this project, a Deep Learning Neural Network has been developed using Matlab, with which (Neural Network), given a picture of a person, the Neural Network can decide and help the doctor's decision about whether the person has BCC or not. The only thing that this software needs is a photo of the person's face with either one of the person's eyes or both of them.In order for the Deep Learning Neural Network to be trained, it has to be given a number of pictures with both pathogenic pictures with faces of persons and pictures of non - pathogenic faces of persons. After the training period, the Neural Network will have its weights calculated. This is necessary, so that the Neural Network can conclude if a photo with a person's face can be considered as pathogenic or not. Parallel to the Deep Neural Network, a Shallow Neural Network was developed and the accuracies were compared. It turns out that the Deep Neural Network can achieve 16better results in terms of the accuracy with which a face of a person (on a photo) can be evaluated to have pathogen or not. Actually, the usual accuracy of the Deep Neural Network is over 80% in evaluating the pathogen in a picture of a person's face. On the other hand, the Shallow Neural Network can achieve evaluation accuracy on the level of about 60% only! This project achieves two things:• Confirms the theory on Deep Neural Networks and Shallow Neural Networks.• Can be a useful tool on helping doctor's decision whether a person has BCC or not
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Chatzopoulos, Evangelos-Georgios
- Χατζόπουλος, Ευάγγελος-Γεώργιος
Subjects
dc:subject × 8Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/49895