Συντάχθηκε 11-10-2013 15:59
από Balasi Panagiota
Email συντάκτη:
Ενημερώθηκε:
-
Ιδιότητα: -.
ΠΟΛΥΤΕΧΝΕΙΟ ΚΡΗΤΗΣ
Τμήμα Ηλεκτρονικών Μηχανικών & Μηχανικών Υπολογιστών
ΠΑΡΟΥΣΙΑΣΗ ΜΕΤΑΠΤΥΧΙΑΚΗΣ ΔΙΑΤΡΙΒΗΣ
ΛΙΒΑΝΟΥ ΓΕΩΡΓΙΟΥ
με θέμα
“ Ανάλυση κυτταρικής εικόνας μικροσκοπίου ”
Τρίτη 15 Οκτωβρίου 2013, 09:30 πμ
Εργαστήριο Προγραμματισμού και Τεχνολογίας Ευφυών Υπολογιστικών Συστημάτων Αίθουσα 141A14 ,Πολυτεχνειούπολη
Εξεταστική Επιτροπή
Καθ. Ζερβάκης Μιχάλης (επιβλέπων)
Καθ. Πετράκης Ευριπίδης
Καθ. Λιάβας Αθανάσιος
Summary
The HER2/neu oncogene is notable both for its role in the pathogenesis of breast cancer and for its selection as a target of treatment. Overexpression of this receptor in breast cancer is associated with increased disease recurrence, poorer relapse-free survival and worse prognosis. Qualitative and quantitative Her2 protein evaluation has been achieved using immunohistochemistry (IHC) on frozen and archival tissues, a widely adopted technique due to its low-cost applicability to any laboratory. When reporting results, the degree of HER2 protein overexpression measured is scored according to the intensity of membrane staining and the percentage of tumor cells stained. The evaluation procedure is usually performed qualitatively by a pathologist, yet the interpretation of such results is subjective and causes certain inconsistencies upon the diagnosis. Our research is focused on the automated detection of Her-2/neu protein in tissues, attempting to make immunohistochemical studies more objective, developing a quantitative technique, based on computer-assisted microscopy and image analysis. The goal of this thesis is to process the extracted IHC microscope images of breast tissues and automatically determine the impact of cancer on the specific female organism. Advanced image analysis techniques are adopted in order to accurately segment the cells within the sample images. Color model conversion enhances the pixel intensity differences between the regions of interest in the test images, clustering reveals the key segments for the evaluation procedure, edge following and linking via the active contours algorithm is performed in order to extract the complete border of the cell membranes and finally the percentage and intensity of their staining is calculated, producing the tissue characterization according to the IHC scoring system. The comparison of the results generated by the proposed algorithm with the conclusions of the classification process performed by the specialist confirms the efficiency and hot prospect of the presented work. Another region discrimination technique combined this time with multispectral polarization imaging technology is implemented. Diffused polarimetric reflection and backscattering provides unique, discriminatory material signatures, based on the depolarization of the impinging waves from different materials. Fusing statistical analysis with polarimetric principles can be proved a powerful tool for analyzing the different properties of operational modalities depicted in digital images. We apply robust fitting of the intensity distributions of the various materials depicted in the acquired images using mixture models and associate the image contrast with the separability of model distributions. Application of this methodology has started to be performed in tissue characterization revealing qualitative discrimination among different types of lung cancer. The chapters in this Thesis present the theoretical study and experimental tests in the previous two areas of tissue characterization at different levels, along with the results obtained and the conclusions generated
