Back to search

National and Kapodistrian University of Athens

Η συμβολή των μεθόδων τεχνητής νοημοσύνης στη διάγνωση αλλοιώσεων του στομάχου

Abstract

dc:description

Objective This study aims to assess the potential value of back propagation artificial neural networks in the discrimination of benign from malignant gastric cells and cases in routinely prepared gastric smears. For purpose of the study, morphometric data were extracted from each cell and were then evaluated by two neural networks. Materials and methods The study was carried out on brushing cytology smears obtained during gastroscopy from 120 patients. The smears were routinely prepared and stained by the Papanicolaou technique. The cytologic diagnosis was made by two experienced Cytopathologists and was histologically confirmed. The result of this procedure was 13300 measured cell nuclei each one being represented by a pre-specified set of 25 geometric and densitometric characteristics. These features were extracted by the image processing system and their values underwent normalization. The entire data set was split into two parts: the training set and the test set. Two different kinds of split were done in order to evaluate the role of dysplastic cells in the networks’ learning procedure. In the first split, data from dysplastic cells were not included in the training set, as was done in the second. Using these data, two different back propagation neural networks were trained and then tested in classifying the cells into two categories (benign/malignant). The endoscopies were performed at the Gastroenterology Unit of the Department of Pathophysiology, Laiko Hospital, while the gastric smears were examined at the Department of Histology and Embryology of the Medical School of Athens University and at the Department of Diagnostic Cytopathology, University General Hospital Attikon. Results Of a total of 13300 measured cells, 2650 were used for training the first neural network and 2300 for the second. The correct classification of these cells reached the percentage of 98-99%. The remaining 10650 or 10300 cells were used for testing the two networks. The correct cell classification percentages by the two BP neural networks were 95,8% and 97,3% respectively. Conclusions The back propagation neural networks gave excellent results in the discrimination between benign and malignant cells from gastric smears. Their performance is improved when data from dysplastic cells are included in their training. The combination of geometric and densitometric characteristics is important in the evaluation of the cells. This method produces significant results in the nuclear and patient level and promises to be a powerful diagnostic tool in clinical practice.

Degree

thesis:*
Grantor dc:publisher
National and Kapodistrian University of Athens
Year dc:date
2010

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Sioulas, Athanasios
  • Σιούλας, Αθανάσιος

Subjects

dc:subject × 14

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/24267

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Sioulas, Athanasios; Σιούλας, Αθανάσιος. Η συμβολή των μεθόδων τεχνητής νοημοσύνης στη διάγνωση αλλοιώσεων του στομάχου. National and Kapodistrian University of Athens, 2010. http://hdl.handle.net/10442/hedi/24267