Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

A Tomographic hyperspectral imaging method: application in quantitative pathology

Chatziioannou Anastasios

Full record


URI: http://purl.tuc.gr/dl/dias/330B6D85-ADC8-4CBA-AD73-D57311C5CED1
Year 2020
Type of Item Master Thesis
License
Details
Bibliographic Citation Anastasios Chatziioannou, "A Tomographic hyperspectral imaging method: application in quantitative pathology", Master Thesis, Σχολή Ηλεκτρολόγων Μηχανικών και Μηχανικών Υπολογιστών, Πολυτεχνείο Κρήτης, Chania, Greece, 2020 https://doi.org/10.26233/heallink.tuc.85673
Appears in Collections

Summary

A hyperspectral imager is a device that records spectra in the form of a spectral cube (a set of spectral images). There are four types of imagers: whiskbrooms, pushbrooms, staring and snapshot. Nowadays, the market is ruled by pushbroom hyperspectral imagers that require spatial scanning. Spatial scanning - as a mandatory requirement - greatly complicates the process of spectral cube acquisition and introduces spectral and spatial distortions. Staring imagers, on the other hand, require no spatial scanning, but suffer from poor spectral resolution and low throughput. This motivated us to build our own staring-type Hyperspectral Imager. The goal was to build a device that requires no spatial scanning and outputs clear spectral images and narrow spectral bands. In this thesis we will uncover how – for the first time – a tomographic spectral cube reconstruction procedure was implemented on a hyperspectral imager. Tomography has never been used in the past in hyperspectral imaging. A separate goal was to design our imager in a way that can be easily adapted on top of digital microscopes.This thesis will explain how a raw spectral cube can be reconstructed to provide high-fidelity spectral cubes with very high spatial resolutions (≥ 6 MP) and very high spectral resolution (wavelength errors ≤ 1nm). This reconstruction procedure involves our own novel Tomographic Reconstruction algorithm that utilizes the whole spectral cube to output the spectrum of every single pixel. We essentially acquire 6 million spectra per band - for 121 bands - a total of 726 million high-fidelity spectra per cube.Our device finds countless uses in GIS, the food and pharmaceutical industries, but our applications are mainly focused in life-saving solutions, like Quantitative Pathology, where multiple stains may be applied simultaneously on top of blood samples and then separated via spectral unmixing.

Available Files

Services

Statistics