Abstract
The use of microwave imaging for medical applications has attracted growing interest in the past decades, inspired by the presence of differences in the dielectric properties of healthy and diseased tissues. The main goal of this research project is focused on incorporating computationally efficient microwave tomography algorithms into medical diagnostic devices.More specifically, microwave tomography methods for head imaging are challenged by the non-linearity of the problem and the non-uniqueness of the solution. The main objective of this thesis is to tackle these challenges and successfully incorporate a previously developed algorithm combining the distorted Born iterative method with two-step iterative shrinkage thresholding (DBIM-TwIST) into a microwave tomography prototype for the problem of brain stroke detection and differentiation.
To this end, we have constructed simplified and complex multi-layer phantoms that have similar dielectric properties with head tissues by developing tissue-mimicking materials formed by gelatine-oil concentrations. In the first set of experiments, the aim has been to show the potential of two-dimensional (2-D) DBIM-TwIST in determining the type of stroke by estimating its dielectric properties. Our results have demonstrated that the prototype can differentiate between hemorrhagic and ischemic strokes based on the estimation of their dielectric properties. The purpose of the second set of experiments has been to experimentally validate our 2-D and three-dimensional (3-D) DBIM-TwIST algorithms for stroke detection and differentiation using a complex head model, but also to further assess experimentally the performance of the 3-D algorithm for the examined application. For some of the studied scenarios, the 2-D DBIMTwIST can lead to more accurate reconstructions. However, the 3-D inverse algorithm can provide more accurate results for problems with significant variation along all three dimensions, as can be the case of brain imaging. This has been the first validation study of a MWI algorithm that detects and differentiates two types of strokes, using an anatomically accurate multi-layered head phantom in a wide frequency range.
| Date of Award | 1 Mar 2023 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Panos Kosmas (Supervisor) & Themos Kallos (Supervisor) |