Quantitative Analysis Of Three Dimensional Vasculature Of Micro-CT Images

Computed tomography has been developed rapidly and received a widespread use in biomedicine since it’s invented by Godfrey Hounsfield and Allan Mc Leod Cormack on 1970 s. It leads to the quantitative analysis of CT images becoming a useful methodology in the research of biomedicine. The abnormality and distortion of the vasculature are commonly seen in many diseases, such as thrombus, tumors and cirrhosis. Therefore, the image analysis after vasculature imaging could be used to extract the quantitative parameters about the morphology and characteristics of the vessel network. These quantitative parameters such as number of vessels, vascular diameter, vessel density etc. have been widely used in the research about the early detection, diagnosis and treatment of related diseases. In recent years, high resolution vasculature images are acquired due to the advanced X ray imaging technique. The outstanding one is the synchrotron radiation based micro-computed tomography(SR-μCT), which characterizes as the high resolution imaging and the achievement of phase contrast imaging(PCI). X ray PCI shows advantage in investigating low Z samples in comparison with the traditional X ray absorption imaging. For example, the PCI is able to detect the biological soft tissues where the density variation between tissues is too small to be detected by absorption imaging. It is normal to use contrast agent for X ray absorption imaging in order to get clear vasculature in CT image. However, the micro vessels are not easy to be captured in this way when contrast agent was perfused into the vasculature. Because the particles of contrast agent are too large to penetrate into many micro-vasculatures. On the contrary, the PCI could acquire vasculature images without contrast agent. Moreover, the micro vessels are also inspected in the CT images. Therefore, the high-resolution 3D images with elaborate vasculature and large image size introduce challenges to the structural analysis and the extraction of quantitative information.This thesis dedicates to investigate the quantitative analytical method of 3D vasculature images acquired by micro-CT in order to satisfy the high demand at the Xray Imaging and Biomedical Application beamline of Shanghai Synchrotron Radiation Facility(SSRF). The main contributions of the thesis are summarized as follows:1. Develop a quantitative analysis method based on the tree-like structure of vasculature. This method traverses the whole vascular tree from the root to extract the quantitative parameter of each vascular element, such as the vessel length, branch points, vessel diameter etc.. Some parameters concerning to the tree structure of vasculature are also extracted, such as the tree level of vessel segment, the highest level of the vascular tree which indicates the degree of vasculature extension. There tree parameters are the supplement of the existing quantitative parameters which are useful for a further investigation of vasculature from different aspect.2. Skeletonization is a vital step for quantitative analysis of vasculature. Generally, it is an effective way to analyze this “compact” representation of vasculature because the structural parameters are readily conveyed in terms of this linear skeleton which is one pixel width. The precision and reliability of the quantitative result are directly related to the accuracy of the skeleton. However, the current skeletonization methods make more effort to preserve topology rather than geometry, which lead to the inaccuracy of skeleton, such as the excessively shortened length of elongated objects, eliminated branches of vessel in tree structures, and numerous noisy spurious branches. In this thesis, we propose a new method to detect the end points of vascular and microvascular branches before extracting skeleton. Then these end points are applied as the constraints in the extraction of the skeleton. The extracted skeleton can preserve the geometry properties of the vasculature and therefore leads to much more accurate quantification of the vasculature.3. The extraction of skeleton always consumes plenty of time, which confines the efficiency of quantitative analysis. By means of Open MP, a parallel designing method based on sequential thinning is proposed to improve the computational time of the skeletonization. The testing results show that the proposed method not only extract precise skeleton, but also conspicuously reduces the processing time to an acceptable scale. For example, the computational time is reduced from 176 min to 13 min for a CT image with a size of 1.95 GB. Therefore, the time efficiency of quantitative assessment is no longer an obstruction for the analysis of the large scale 3D vascular CT images.4. The quantitative method and the improved skeletonization method proposed in this thesis are successfully implemented as a software tool by C++ programming. They are implemented as modules running in 3D Slicer. A brief instruction is presented to show the workflows of analyzing vasculature in a 3D image.5. The quantitative analysis method proposed on this thesis is introduced to investigate the mouse model of liver fibrosis, by which the degree of hepatic fibrosis is assessed. A new parameter is extracted from the quantitative result of our vascular tree analysis.The experimental results demonstrate that the new parameter(HVR) is robust and reliable to evaluate the degree of hepatic fibrosis, which means that the developed method is practical for the evaluation of liver fibrosis.