Validation and qualience-mediated computed tomography indeep tissue regions
Rosenhain, Stefanie; Kießling, Fabian (Thesis advisor); Strnad, Pavel (Thesis advisor)
Aachen (2019, 2020)
Dissertation / PhD Thesis
Dissertation, RWTH Aachen University, 2019, Kumulative Dissertation
Fluorescence imaging is a well-established and widely used technology in many research areas. Fluorescence images can be acquired even in 3D by using confocal and 2-photon microscopy, but only in a very low depth of around 100 to 200 µm. An improvement in the field of fluorescence imaging was the invention of fluorescence-mediated tomography (FMT). FMT imaging enables a non-invasive, three-dimensional, and longitudinal assessment of the distribution of fluorescent molecules for example in the whole murine body. However, FMT imaging was practically restricted to superficial structures for a very long time, because the irregular mouse shape and heterogeneous absorption and scattering coefficients of biological chromophores such as hemoglobin, melanin, and fat, hamper a correct fluorescence signal reconstruction in deep regions. This obstacle was overcome by combining FMT with an additional imaging modality, computed tomography (CT), that provides anatomical information and promotes a more accurate organ segmentation. Nevertheless, the validation and quality assessment of hybrid fluorescence-mediated computed tomography was and is still challenging. For validation purposes, plastic-, agarose-, gelatin-based or 3D-printed phantoms are intensively used. However, they are not representative for in vivo experiments since the specific absorption and scattering coefficients of biological tissues depend on oxygenation state, are highly heterogeneous, difficult to measure, and thus, hard to be simulated with phantoms. Therefore, the aim of my thesis is to validate hybrid FMT-CT imaging especially in deep tissue regions and evidence its suitability for preclinical imaging studies. In my first publication, I established an in vivo protocol based on rectal insertions containing specific amounts of fluorescent dyes to assess the sensitivity and accuracy of 3D fluorescence imaging. Thereby, I demonstrate that the improved FMT-CT reconstruction algorithm allows an accurate detection of fluorescence signals in deep tissue regions even in a picomolar range. Thus, preclinical imaging studies can be performed to determine fluorescence intensities in the entire body. As part of these studies, a huge number of FMT-CT data are generated through multiple groups, animals, probes, and points in time. In order to assess these images, single organs or regions need to be segmented, e.g. by manual delineation. However, this is a time-consuming, error-prone, and user-dependent approach with a low reproducibility. To facilitate the development of more reliable automated segmentation algorithms for future FMT-CT studies, I introduce a CT database consisting of native and contrast-enhanced CT scans along with 3D organ segmentations in my second publication. By performing over 225 whole body organ segmentations and calculation of the Sørensen-Dice coefficient, I demonstrate high user-dependent errors due to manual analysis especially for liver and spleen. In summary, both publications together contribute towards establishment of FMT-CT imaging as an accurate, robust, and efficient tool for basic and applied science.