Extraction And Tractography Of Brain White Matter Region

Human brain is one of the most complicated systems in the world, and the exploration in the mechanism of human brain information processing is also one of the most chanllenging problems among science reserch. The development of modern imaging technique makes the research of human brain in a non-invasive way become available. From previous research based on EEG, MEG and fMRI techniques, it indicates that human brain has different functions in differnt regions, which means "functional segregation". However, even when the human brain implements an extremely process, it always involves many different regions interacting and cooperating with each other, thereby constructing a network to complete the task, and this is called "functional integration". As the human brain can be considered as a very complicated network, it is necessary for us to investigate the function of human brain on the basis of network.There were numerous papers about the construction of brain network in recent years, which makes this topic become the most welcomed one among science research. The essential step in the process of constructing the brain network is tracing the white matter bundles, which is called tractography. Therefore, accurate tractography is of critical significance for efficient brain network construcation. So far, there are numerous tractography techniques, and they can generally be classified into two catigories: deterministic tractography and probabilistic tractography. Traditional deterministic tractography is very fast but inaccurate in regions where fibers cross or twist within the voxel. Probabilistic tracking methods are accurate but a time-consuming process and difficult to interpret, making the clinical use unavailable. Therefore, this thesis focused on investigating the latest techniques of tractography and exploring the application of tractography. The major contributions in this thesis are the following points:1. Considering both the accuracy and speed of the algorithm, in this thesis we proposed a combinatorial method based on a two-tensor model. As the two-tensor model is able to address the fiber crossing problem, it will improve the accuracy of the algorithm. Also the deterministic method is very fast so it is possible to decrease the computational time. The proposed Combinatorial Streamline Tractography (CST) is a tradeoff for speed and accuracy, so that the tractography can be used in clinical practice more efficiently and accurately.2. Simultaneously, in order to stop the white matter bundles from tracking outside the white matter (WM) region, we extracted the WM region by using an improved Random Walks method and achieved a binary template, thereby limiting the white matter bundles in the WM region. However, due to the complex anatomical structure of brain tissue, the original Random Walks cannot achieve a good extraction result. In order to achieve a more accurate WM region, we introduced Local Binary Pattern (LBP) and Prior Probability Model to Random Walks to improve the accuracy of this algorithm.3. We evaluated the performance of our proposed Combinatorial Streamline Tractography both on synthetic datasets and real brain diffusion MRI datasets. The results demonstrate that this approach not only successfully reveals structure in crossing regions over a broad range of crossing angles and curvatures, but also it is efficient and robust for clinical use.This research was sponsored by Nature and Science Foundation of China (Project No.60771007) and Chinese Academy of Science Graduate Innovation Funding.