Application Research On Medical Sequential Images Fast Clustering And CT Image Reconstruction

Cluster analysis or clustering is the task of assigning a set of objects into groups (called clusters) so that the objects in the same cluster are more similar (in some sense or another) to each other than to those in other clusters. It has significant clinical value.In recent years, with the improvement of diagnostic imaging requirements and the development of medical imaging technology, most of medical images exist in the form of sequential ones in clinical. Classic image clustering methods directly applied in medical image sequences has many problems. At present, aiming at medical image sequences cluster analysis is still in the early stages of study, and it makes the research of relevant clustering methods is of great significance. In this thesis, we analyzed the traditional clustering algorithms deeply, then we put forward two kinds of fast clustering methods for medical sequential images. One was based on the template and another was based on linear transform and BP neural network. The experimental results showed that these two algorithms not only had high accuracy, but also had high speed. We believe the two methods will have good prospect.On the other hand, in recent years, with the development of computer tomography (CT) technology, it’s widely used in clinical. But in practical applications, it often subject to some objective restrictions or sometimes we should consider some special clinical application requirements, so that the projection data we get can’t meet the data completeness conditions. So study on how to accurately reconstruct images from insufficient projection data not only have theoretical values, but also have clinical values. There are a lot of popular image reconstruction algorithms, for example, the filter backprojection (FBP) algorithm which has more extensive application, but FBP algorithm requires sufficient projection data; In addition to, there are a lot of iterative image reconstruction algorithms, which have large computation but can deal with the insufficient projection data. Therefore, the iterative image reconstruction algorithms are the best choice to solve insufficient projection data problem.In the thesis, we analyzed the properties of the traditional reconstruction algorithms deeply, then we further classified the image reconstruction algorithm according to the different ways to calculate the projection and backprojection, i.e., pixel-driven, ray-driven and distance-driven CT image reconstruction. In this thesis, we implemented these three image reconstruction algorithms and analyzed the performance of these algorithms. Finally, we developed an iterative image reconstruction algorithm which could deal with the insufficient projection data. This algorithm was based on the distance-driven reconstruction technique and the total variation optimization (DD-TV algorithm). The simulation experimental results showed that DD-TV algorithm has higher application value.