Institutional Repository
Technical University of Crete
EN  |  EL



My Space

Novel spectral mapping indices for improving microscopy and leukemia diagnosis

Kastrinakis Marios

Full record

Year 2017
Type of Item Master Thesis
Appears in Collections
Relations with other Items


The rise and evolution of optical light microscopes is inevitable, as microscopy constitutes a key diagnostic and imaging tool for many fields of science, such as medicine, biology, chemistry as well as for many applications in industry. However, the internationally accepted technology residue regards to the physical limitations of the microscopes’ resolution and the biological limitations of the examiners’ perception, quest the improvement of imaging ability in microscopy. In addition, the evaluation of the imaging information of diagnostic importance urges the development of computational methods that will highlight and measure it quantitatively.In this thesis, a novel method is reported for measuring and enhancing imaging resolution in microscopy, based on a single measurement of one-dimensional modulation transfer function combined with Wiener deconvolution algorithm. Additionally, for the first time in the relevant literature to the best of our knowledge, a computational method is presented for improving diagnostic accuracy of leukemia, using optical light microscopes along with spectral imaging. A set of spectral indices are introduced that have been extracted and measured quantitatively, through chemometrics regression algorithm combined with spectral imaging of peripheral blood smears.The proposed methods are capable to achieve enhanced optical resolution of up to 176% for a brightfield microscope at 23.8x of total optical magnification, beyond microscope’s optical limit. Furthermore, they achieve 100% sensitivity with 98.91% specificity in cases of Acute Lymphoblastic Leukemia, by detecting lymphoblasts accurately and 74.86% sensitivity with 96.94% specificity in cases of Chronic Lymphocytic Leukemia, by differentiating and indicating abnormal lymphocytes from the normal ones quantitatively. Thus, the proposed methods can be easily employed as additional analytical tools for minimizing errors and increasing accuracy, either in the diagnosis and in the classification of ALL and CLL leukemia.

Available Files