Back to search

Aristotle University Of Thessaloniki (AUTH)

Ανάλυση ιατρικών δεδομένων από πολλαπλές απεικονιστικές τεχνολογίες με τεχνικές μηχανικής μάθησης, επεξεργασίας εικόνας και στατιστικού μοντέλου σχήματος

Abstract

dc:description

This thesis describes innovate methods of image processing on data having been collected through various medical imaging systems (OCT and SPECT / CT) and related to specific issues - problems of the diagnostic medicine. These methods were developed in accordance with the principles of image processing, by using convolutional neural networks and active shape models. We examined basically images of Intravascular Optical Coherence Tomography (which constitutes a relatively new and evolving medical image technology), in view of the morphology and pathophysiology of the vascular arteries, and with the aim of developing methods for automatic segmentation of the arterial wall and classification of atherosclerotic tissues. We developed the ARC-OCT method which enables accurate and fully automatic detection of lumen-endothelial borders even in OCT images containing artifacts, arterial stented segments and lateral branches. For the automatic classification of atherosclerotic plaques, (after having apply - test a variety of alternative techniques) we developed a combinatorial method based on the specialized pre-treatment of OCT images (to highlight the optical properties of the different types of atherosclerotic tissues) and classification of them in two level bases: textural and regional, using the appropriate CNNs network at each level. The method was used to classify portions of an IVOCT image into 6 categories: background, normal tissue, and four different types of atherosclerotic plaque (lipid, fibrous, calcified and mixed). Furthermore, for the automatic segmentation of V/P SPECT functional images of the lungs - another field of our research work- we developed a method which is based on Active Shape Models, that is the transferring the a priori information, derived from the corresponding anatomical CT, to the segmentation procedure. The algorithm that was developed allows estimating the contour of the lungs independently of the existence of peripheral ventilation / perfusion defects and can form the basis for the automatic quantification of large-scale of relevant measurements. Finally, we identified issues (on the above fields) that require further research and development, that can be done by using deep learning methods and 3D representation techniques, so that the results of our research effort can be more widely applied in clinical practice.

Degree

thesis:*
Grantor dc:publisher
Aristotle University Of Thessaloniki (AUTH)
Year dc:date
2020

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Cheimariotis, Grigorios-Aris
  • Χειμαριώτης, Γρηγόριος-Άρης

Subjects

dc:subject × 14

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/46981

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

Cheimariotis, Grigorios-Aris; Χειμαριώτης, Γρηγόριος-Άρης. Ανάλυση ιατρικών δεδομένων από πολλαπλές απεικονιστικές τεχνολογίες με τεχνικές μηχανικής μάθησης, επεξεργασίας εικόνας και στατιστικού μοντέλου σχήματος. Aristotle University Of Thessaloniki (AUTH), 2020. http://hdl.handle.net/10442/hedi/46981