Institutional Repository
Technical University of Crete
EN  |  EL



My Space

Multiband, dynamic and molecular imaging

Giakoumakis Theodoros-Marios

Full record

Year 2016
Type of Item Master Thesis
Bibliographic Citation Theodoros-Marios Giakoumakis, "Multiband, dynamic and molecular imaging", Master Thesis, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2016
Appears in Collections


Tumour heterogeneity describes differences of tumour cells either within the same malignant mass or amongst patients suffering from the same cancer type. It has only recently begun being explored by modern medical science and the repercussions in effective screening and treatment have yet to be fully understood. While research is progressing in search of new biomarkers, there is an increasing tendency to use multiple biomarkers to achieve more precise diagnosis. Under this light, we present a novel multiband bio photonic method and imaging modality capable of imaging in vivo dynamic phenomena and estimating and mapping functional and structural parameters of the observed tissue, fully customizable to record a given set of biomarkers. Complete spectral cube of dynamic phenomena, induced by multiple biomarkers, can be achieved by employing spectral estimation algorithms and single frame simultaneous acquisition of a limited number of spectral bands. The pharmacokinetics of each biomarker can be algorithmically separated, with use of regression algorithms, based on their spectral signature. In silico modeling of the pharmacokinetics of these biomarkers, using compartmental analysis, may adopt sets of biological parameters correlated with the observed response. Previous studies have reported a model of the cervical neoplastic epithelium which predicts the dynamic optical effects after the application of acetic acid. Global optimization analysis has proven that four structural and functional biological parameters correlated with cervical neoplasia can be accurately estimated, with the Differential Evolution (DE) algorithm achieving roughly 99% mean accuracy, as the best performing one. In the scope of this study, we proceed to estimate and map thousands of parameter sets from clinical data, using either DE or a look-up table (LUT) consisting of predefined solutions with at least 92% mean accuracy. The findings agree with literature. Moreover, the use of LUT is vastly time efficient compared to the use of DE, offering estimations within a couple minutes instead of days with similar accuracy levels. The suggested imaging modality can be proven an indispensable tool for reducing time and cost of screening procedures, by providing accurate, quantitative and objective information of biological parameters after the application of multiple biomarkers, while on the same time can be used to further understand dynamic phenomena through real time spectral imaging.

Available Files