Back to search

National Technical University of Athens (NTUA)

Μελέτη - ανάπτυξη και αξιολόγηση ολοκληρωμένου συστήματος ψηφιακής ανάλυσης ενδοσκοπικής εικόνας μακροβιοψίας για την υποστήριξη διάγνωσης γυναικολογικού καρκίνου

Abstract

dc:description

Cancer is the second cause of death in the USA with over 557,271 deaths in 2002¹ based on data from the National Center for Health Statistics. The gynaecological cancer is among the second cause of death and according to the American Cancer Society in 2005 will be diagnosed over than 40,880 females with cancer. Also, according to the same source the deaths from uterus cancer will increased upon 7,310 in 2005² in the USA. Laparoscopy has become the preferred surgical technique for several medical cases such as gynaecological cancer because it is minimally invasive. The physician guides the camera inside the human body in order to investigate suspicious areas of cancer. In difficult cases of gynaecological cancer the physician can’t classify the suspicious area and the need of histopathological examination is necessary. The laparoscopy/hysteroscopy examination provides direct information of the status from the female reproductive organs. Especially, hysteroscopy is a routine examination for the endometrium, fallopian tubes, uterus and the body of the uterus. The main objective of this study was to develop a Computer Aided Diagnosis system (CAD) to support the physician in difficult cases of gynaecological cancer for the identification of suspicious areas of cancer. Sometimes the identification of Regions of Interest (ROIs) for gynaecological cancer could be a difficult process. The need of a CAD system could be a useful tool in the hands of the physician. To the best of our knowledge no similar study was performed for laparoscopy imaging. Similar work was carried out only for colonoscopy images for the detection of tumours in colonoscopic video where the performance in the detection of abnormal regions corresponding to adenomatous polyps was high. Our research consists of 4 main procedures. In first procedure we developed a database for gynaecological cancer for laparoscopy/hysteroscopy cases. We collected 416 images of endometrium from a total of 40 subjects. The second main procedure involves the development of a complete system of laparoscopy/hysteroscopy image analysis. This system consists of the following modules: (i) Capturing of video: The output signal of the laparoscopy camera (PAL 475 horizontal lines) was digitized using a frame grabber and was saved in a personal computer. The video was saved with a resolution of 720x576 pixels x24 bits (AVI format). The physicians freeze the images (720x576 pixels x24 bits Tiff format) and the ROIs were segmented manually, (ii) Gamma correction algorithm was applied for the calibration of the medical camera. The meaning of the gamma correction was to standardise the capturing protocol facilitating the quantitative analysis. Regions of Interest (ROIs) were segmented manually by the physician, (iii) Textures features based on statistical, Spatial Gray Level Dependence Matrices (SGLDM), Statistical Features (SF) and Gray Level Difference Statistics (GLDS) and color histograms of ROIs of suspicious areas of cancer were computed. Experiments with different viewing conditions were investigated. Results showed that in small distances and in near angles there is no statistical difference between the ROI texture features. In the cases of endometrium, results shown that most of the features have statistical difference between (Wilcoxon Rank Sum test, a =0.005) normal and abnormal ROIs. Features like the mean, entropy have significant statistical difference, (iv) Multiscale analysis was applied in all cases beginning from 1x1 until 10x10 scales. The images were compared statistically in all scales. At scales above 4x4, images were destroyed. In the third procedure neural network algorithms were applied for the classification of normal and abnormal ROIs. Results showed that suspicious areas can be identified in normal and abnormal ROIs using the system. In a hyperplasia case the percentage of correct classifications was 74% in normal ROIs and 83% in abnormal ROIs respectively, with maximum correct classification percentage 79%. Finally the last procedure, number four, involves the results of our research and the evaluation of the system with the marketing approach. The results were very encouraging in forwards the development of a commercial system CAD system for gynaecological cancer. Our system was proposed to be used by the European Society for Gynaecological Endoscopy (ESGE) and the European Academy of Gynaecological Cancer (EAGC) for the analysis of endoscopy images. Also, the results of the research work was published in 11 international conferences and 2 in international scientific journal but also in the near future, 2 more papers will be submitted in conferences and in scientific journals.

Degree

thesis:*
Grantor dc:publisher
National Technical University of Athens (NTUA)
Year dc:date
2007

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Νεοφύτου, Μάριος
  • Neofytou, Marios

Subjects

dc:subject × 20

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/16875

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

Νεοφύτου, Μάριος; Neofytou, Marios. Μελέτη - ανάπτυξη και αξιολόγηση ολοκληρωμένου συστήματος ψηφιακής ανάλυσης ενδοσκοπικής εικόνας μακροβιοψίας για την υποστήριξη διάγνωσης γυναικολογικού καρκίνου. National Technical University of Athens (NTUA), 2007. http://hdl.handle.net/10442/hedi/16875