Optimization of μct imaging systems for soft-tissue dedicated tasks
- Autores: Ana Ortega Gil
- Directores de la Tesis: Juan José Vaquero López (dir. tes.)
- Lectura: En la Universidad Carlos III de Madrid ( España ) en 2019
- Idioma: español
- Tribunal Calificador de la Tesis: Juncal Garmendia Garcia (presid.), Javier Pascau González-Garzón (secret.), Mathieu Dupont (voc.)
- Programa de doctorado: Programa de Doctorado en Ingeniería Eléctrica, Electrónica y Automática por la Universidad Carlos III de Madrid
- Materias:
- Enlaces
- Tesis en acceso abierto en: e-Archivo
- Resumen
Tuberculosis tends to be seen as a disease of the past, however by reviewing this infectious disease and its prevalence it arises the need for new drug combinations for the treatment of drug-susceptible microorganisms proliferating. The experimental requirements for the development of new drug combination regimens are imposed by the experimental animal model and the biosafety requirements intrinsic to infectous disease research with Mycobacterium Tuberculosis (Mtb).
Despite the efforts, the critical development path (preclinical phase) can take more than 6 years to enter the human clinical testing phase. By introducing imaging tools, the preclinical phase can be shorten and the success probabilty of the candidate drug combination can be increased. Imaging allows to follow up the disease progression with mild suffering of the animal and every animal can reach the humane end points stablished at the assay design. At those endpoints, histopathology will confirm and extend the information already seen in the images. Thus preclinical imaging complies with the guiding principles of the 3Rs (replacement, reduction and refinement) for performing more humane animal research.
This thesis is focused on the refinement of soft tissue X-ray imaging in small animal micro computed tomography (μCT) for assessing the host response of pulmonary infectious diseases in the preclinical stage of drug combination development programs, i.e. in in vivo efficacy assays. We propose refinement interventions at every application of the imaging system: at the acquisition, by optimizing the protocol for soft-tissue imaging tasks; at the proccessing level, by quantifying the disease progression with a novel radiological biomarker; and at the histological analysis, by guiding the block slicing straight to the lesions of interest, previously located in the μCT of post portem samples with metallic staining.
The μCT technology has the advantages of being a highly developed tool for hard-tissue dedicated tasks and of allowing periodical high-troughput imaging acquisition of the experimental subjects. However, to over come the artifacts in soft-tissue imaging, we propose a first refinement strategy at the system design.We ensambled a new μCT system with the latest detection technology and conceived it as a flexible cone-beam platform (CBμCT) with variable parameters, permitting the adaptation of the geometrical configuration to assessing lung infection burden.
The optimization of the system calibration and working point corrected the artifacts due to the low contrast inherent to soft-tissue density and to the electronics and mechanics. The avoidance of the respiratory motion artifact is also achieved by the design of dedicated acquisition protocols to stop inconsistencies propagation. Additionally, an image processing algorithm is proposed for sorting projections into respiratory phases. The optimized protocol has a beam configuration set at 68 kV (x-ray source voltage) and, 420 uA (anode current) with a soft-tissue filter (0.2 mm Cu + 1.8mm Al, intinsic) . We selected a 0.7° step-and-shoot protoco covering 360° with a pixel binning of four-by-four pixel binning, and a multi-frame rate of four frames per gantry position at 20 frames per second. These acquisition parameters lead togenerate a total data volume data of 2.17GB (1.44MB per frame) and a total acquisition time of eight minutes. Data-sets were reconstructed using the filtered back-projection (FBP) algorithm and an isotropic voxel resolution of 88 um.
The prototype has total auto shielding for being host in any room and the metallic components and the cover itself are resistant to the disinfection protocol by air with VH2O2 and with UV light., to be compliant with the biosafety-level-3 measures for working with Mtb bacillus, considered as a level three pathogen agent with high individual risk and low population risk. An animal cartridge is also designed .as primary biosafety barrier. The physical container has an anaesthetizing environment of a mixture of anaesthesia/air-oxygen continuously circulating so when it is hermetically closed with the animal mounted, it can be safely manipulated in the equipment room and positioned in the imaging system. Once the quality of the images reaches the task requirements, the system is placed in production and μCT imaging is included in the efficacy assay planning. For the complete characterization of the disease progression, imaging tasks are included in both the in vivo and post portem stages of the experimental desing .
The second refinment strategy focuses on the quantification of the disease progression by imaging. We design, validate and stablish an in vivo radiological involvement score as the biomarker for the assesment of differential virulence and disease progression. The dataset used for the design included scans of the disease progression of two mouse strains, infected with two different virulent Mtb.
The extraction of the radiological biomarker consists on the segmentation by intensities of the thoracic region and the quantification and normalization of the connected uninvolved tissue in that region. From the μCT scan of a mouse in the isolation cartridge, the mouse contour is identified. Then, the uninvolved lungs are segmented. The μCT biomarker is computed as the normalized segmented uninvolved lung volume.. The premise of finding connected components when segmenting uninvolved lung lobes is hold for tuberculosis damage since the infection starts on the alveoli. We decided to include the airway walls in the structural baseline of the CT biomarker because the disease progresses towards collapsing the airways, thus reducing the total volume of the uninvolved lung volume.
The radiological biomarker uses the remaining healthy volume as an indicator of the disease progression, A decrease on the biomarker value can be interpreted as an indication of increasing adverse effects, which could be considered by investigators when developing scoring sheets for humane endpoints and complement well-known objective criteria as the weight decrease. Furthermore, the biomarker has the potential to reduce the sources of bias in prospective studies such as the loss of individuals to follow up due to unsuccessful infection. Immediacy of the analysis results allow faster efficacy assessment, that is crucial for determining the therapeutic potential of new combination regimens for drug-resistant tuberculosis strains.
Framed in efficacy assays, the analysis of 228 biomarker results has characterized the disease progression of three animal models: C-57 mouse infected with 1000 CFUs of H37Rv, C3HeB/FeJ mouse infected with 100 CFUs of TMC107 and C3HeB/FeJ mouse infected with 10000 CFUs of H37Rv microorganisms. Owing the differential virulence of the Mtb strains and of the triggered immune response, the unsupervised classification is described per experimental design. The classifiers will prevail for new efficacy assays introducing new treatments. In general, the disease involvement progression presented minor lesions three weeks after the infection which evolved into severe damage of the lungs, with visible granulomas, within five weeks. The correspondence between the renders of the uninvolved lung from the last micro-CT scan and the excised organs was confirmed and the relative uninvolved values are confirmed by the histological evaluation.
Statistically significant difference among the mean relative uninvolved volume of micro-CT scans of treated and non-infected groups and infected group was confirmed by t-test (p-value < 0.05). The 95% confidence interval for the difference is [2.91, 21.61] and [-23.81, -17.84], respectively. Nevertheless, the differences between the non-infected and the treated groups are not statistically significant (p-value=0.99), with a 95% confidence interval for the difference is [-6.21, 6.53].
As a third refinement strategy, we exploite X-ray imaging capabilities for post mortem lesion characterization and stratification. Tuberculosis histopathology on μCT is achieved by tissue staining with non-specific contrast agents (Iodine and Silver) which tend to adhere to highly cell populated regions such as granulomas. The contrast increase due to the metallic staining provides detailed images of the lung structures and tuberculosis lesions, which is an advantageous information for 3D histology in whole organs and for guiding histological slicing. Post mortem μCT also allows the classification of granuloma intensities by cellular composition, which sets the basis for assessing drug efficacy in terms of penetration by μCT.
We have demonstrated the feasibility of post mortem whole lung silver and iodinated staining for X-ray histology to assess the 3D characteristics of abnormal bronchoalveolar tissue. We showed that this approach is not limited to a single staining agent by presenting comparable results with both solutions. In addition, we demonstrated that our preparation for staining is compatible with disinfection protocols and allows subsequent analysis using classical histology. Our protocol does not introduce any additional shrinkage of the tissue other than that expected from the chemical drying procedure and paraffin embedding typical of standard histological examinations.
Comparison of lung volumes throughout the process shows the expected shrinkage and no shrinkage due to the contrast agents applied. We recommend that samples are stored in 70% ethanol to avoid further volume changes. In the case of paraffin storage, preservation can be assured and moreover, the lungs can be scanned with a slight decrease in the image contrast while continuing to allow planning of the histological cuts to the regions of interest.
Both silver and iodinated contrast agents demonstrate low protein binding, thus explaining the clear depiction of vessels, airways, connective tissue, and lesions. However, silver nitrate produced higher contrast at a 3% concentration (saturated) than iodine at the same concentration on the analysed μCT images, thus facilitating the depiction of the rim and central regions of the lesions.. On the other hand, the iodinated contrast agent is a well-established technique and can be translated to in vivo scans using its injectable form.
The weaknesses of this approach arise from the fact that histological annotations based in intensity have misidentifications owing that multiple tissues are coded with the same colour range. In our case, the characteristic cells composing the walls of collapsed alveoli, trachea and airways are automatically annotated as diseased tissue, both in the 3D micro CT volume and in the 2D histological slide. Spots in the digitalized histology slide due to dust in the lens were also detected by the classifier as diseased tissue. Furthermore, the staining concentration also interferes the classification, preventing the extraction of a global set of thresholds for CT slices or histological slides.
It must also be considered that there is not exact correspondence between the histology image and a slice in the μCT. The voxel size of our 2D μCT slice is 44 μm, meaning that the image is a flat cross-section which integrates all the structures within 44 um depth, whereas the width of the histology samples is 3 um. This may lead to misclassification of lesions or patterns that appear in the histology slide but not in the CT slice. One common event is the edge effect. Edges in histology have high intensity values and sometimes classified as neutrophil focus. However, those patterns are not expressed in μCT slice.
Nevertheless, histopathology enabled us to perform a detailed CT examination of the lung anatomy and composition from the payload carried by the contrast materials. Our results confirm that the proposed histopathologic-guided biomarker extraction provides a satisfactory estimation of the volumes of interest by statistically modelling the HU distribution and registering the CT with histological slides. These methods automatically assign thresholds to 3D micro CTs for revealing the presence, extent and appearance of the immune responses to a tuberculosis challenge. Consequently, this method may enable stratification of abnormal CT densities associated with a wide variety of pulmonary diseases in preclinical research: any other model of injury, infection or tumorigenesis which also present clusters of cells as hallmarks of the host response, could be enhanced by metallic stains in μCT imaging.
Cutting-edge spectral systems are expected to demonstrate superior image quality, even with larger voxel sizes than standard CT systems, provided that the attenuation of subjects under study is enhanced. Thanks to the proposed staining protocol, we lauched a proof-of-concept on lung inflammation and infection imaging by preclinical spectral systems.
The state-of-the-art energy -resolving detection technologies for medical imaging, originally developped by CERN and mounted in the prototypes PIXSCAN-FLI (IMxGAM, Fr) and μCT-MARS (MARS Biomimaging Ltd., NZ) were the selected systems to test the potential on infected pulmonary imaging. We determined that they are able to differentially asses uptake of non-specific contrast agents (Iodine and Silver) into normal parenchyma and into Mtb-infected sites in post mortem whole lungs. The current technical limitation for energy-resolved image in the d PIXSCAN-FLI system is the image quality sensitivity to differences on pixel responses, which result in homogeneities in the projection of uniform regions and ring artifacts in the reconstructed sets. In the case of the μCT-MARS, the technical limitation for in vivo is the long acquisition time (~40 min), inviable for high-throughput assays or subjects with low chance of survival of long anaesthesic induction due to the severe affection. An other aspect to take in consideration in spectral imaging is the high dependency of accurate identification/quantification on the size, shape and concentration of the metallic atoms accumulated within the tissue of interest. Many research efforts are nowadays focused on nanoparticle engineering and in light of the contributions of this thesis, future research lines will use specific contrast agents with translational potential to clinical imaging for the stimation of MTb involvement. The ability to make this assesment at imaging would allow early diagnosis by a single scan and personalised treatment of lung infections.
The contibutions in soft tissue X-ray imaging in small animal micro computed tomography (μCT) achieved with thesis are a major refinement of the technolgy, allowing the assessment of disease involvement in the lungs, with translational capabilities for clinical diagnosis of any pulmonary diseases causing lung opacities (i.e injury, infection or tumorigenesis). In preclinical imaging, it significantly reduces heterogeneity and animal numbers while improving reproducibility of disease models which are key issues in non-humane primate models. These outcomes could eventually help bring more effective drugs to patients by improving the predictive power of preclinical methods.
