Back to search

University of Patras

Ανάπτυξη νέων μεθόδων ανάλυσης ψηφιακού σήματος και εικόνας και αναγνώριση προτύπων σε ιατρικές εφαρμογές

Abstract

dc:description

Objective: The objective of the present PhD Thesis was the development and application of image analysis and pattern recognition methods, using histopathology images of laryngeal tumors which had been immunohistochemically (IHC) stained for p63 expression, for: a) investigating the association between P63 stained nuclei and histologic grade in laryngeal tumor lesions, b) designing a decision support system (DSS) for discriminating between low and high grade laryngeal cancer cases, and c) developing DSS-system for predicting the 5-year survival of patients with laryngeal squamous cell carcinoma.Material and Methods: Biopsy specimens from laryngeal tumor lesions of 55 patients diagnosed with laryngeal squamous cell carcinomas were Hematoxylin & Eosin (HE) stained for histological tumour grade and stage assessment, and immunohistochemically (IHC) stained for P63 expression. Four non-overlapping microscopy images were digitized from each patient’s IHC specimens, in regions indicated by the expert physician. In each image, P63 positively expressed nuclei were identified, by specially designed segmentation method, and the percentage of P63 expressed nuclei was computed. A number of textural and morphological features were generated from each of the P63 stained nuclei and data thus formed was used to form three classes (21 lesions of grade I, 34 lesions of grade II and grade III). Features of high discriminatory power amongst the classes were determined, employing non-parametric statistical methods, and the effect of tumour grade on the nucleus structure was analysed. Next, two DSS-systems were developed, employing the CUDA parallel programming capabilities of Nvidia GPU cards for achieving high precision designs. One DSS was designed to discriminate low from high grade laryngeal cancer lesions and another to predict 5-year survival of patients with laryngeal cancer. Results-Conclusions: Textural features that sustained statistically significant differences between low and high grade lesions revealed that with advancing grade, %P63 expression decreased, P63 stained nuclei appeared of lower image intensity, more inhomogeneous, of higher local contrast, containing smaller randomly distributed dissimilar structures and having more irregular shape. DSS performance in classifying laryngeal cancer cases as low grade and high grade was 85.7% and 94.1% respectively. The system’s overall accuracy was 90.9% and its estimated accuracy to “unseen” by the system cases was 80%. The DSS accuracy in assessing the 5-year survival of patients with laryngeal cancer was 90.5%, and texture in p63 expressed nuclei appeared coarser and with more edges in the 5-year non-survivors.

Degree

thesis:*
Grantor dc:publisher
University of Patras
Year dc:date
2015

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Ninos, Konstantinos
  • Νίνος, Κωνσταντίνος

Subjects

dc:subject × 6

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/36311

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

Ninos, Konstantinos; Νίνος, Κωνσταντίνος. Ανάπτυξη νέων μεθόδων ανάλυσης ψηφιακού σήματος και εικόνας και αναγνώριση προτύπων σε ιατρικές εφαρμογές. University of Patras, 2015. http://hdl.handle.net/10442/hedi/36311