Συντάχθηκε 31-03-2014 13:21
από Balasi Panagiota
Email συντάκτη:
Ενημερώθηκε:
-
Ιδιότητα: -.
ΠΟΛΥΤΕΧΝΕΙΟ ΚΡΗΤΗΣ
Σχολή Ηλεκτρονικών Μηχανικών και Μηχανικών Υπολογιστών
ΠΑΡΟΥΣΙΑΣΗ ΔΙΠΛΩΜΑΤΙΚΗΣ
ΚΛΑΠΑΚΗ ΣΤΥΛΙΑΝΗ
με θέμα
Συγκριτική αξιολόγηση μεθόδων καθολικής βελτιστοποίησης για την παραμετροποίηση και ανάλυση βιο-οπτικών σημάτων διαγνωστικού ενδιαφέροντος.
Comparative evaluation of global optimization methods for the parameterization and analysis of bio-optical signals with diagnostic importance
Τρίτη 1 Απριλίου 2014, 13.00 μμ.
Αίθουσα 2042
Εξεταστική Επιτροπή
Αναπληρωτής Καθηγητής Μπάλας Κωνσταντίνος (επιβλέπων)
Αναπληρωτής Καθηγητής Μπλέτσας Άγγελος
Καθηγητής Σταυρακάκης Γεώργιος
Abstract
Dynamic Spectral Imaging (DSI) is a novel imaging modality based on the in vivo measurement of spatial-spectral and dynamic optical characteristics induced by biomarkers. A compartmental pharmacokinetic model of the cervical neoplastic epithelium has been developed, with the aid of which the experimental data are fitted, in an attempt to correlate the in vivo measured dynamic back-scattering (DBS) signal with the actual functional and structural characteristics of the tissue. The eight identified well correlating with neoplasia parameters, were reduced to five with the aid of Global Sensitivity Analysis. In this paper we used five state-of-the-art global optimization techniques, equally constrained in a parallel working software environment in order to verify whether the solution of the inverse problem is substantially unique. The statistical results confirmed that for the reduced set, the unidentified experimental optical data converge more optimally, in comparison with the case where the full set is used. Moreover the differential evolution’s algorithm error is under 10% for the five most identifiable parameters, fact that sets this one as the best performing algorithm of our study and the results quite promising on our way towards the unique estimation of those parameters, indicating a good accuracy and precision.
