Application Of Compressive Sensing In Medical CT Image Reconstruction

Since Hounsfield established the first medical CT in1967, computed tomography with its high resolution, the distinct of anatomical relationship and the good clear show of lesions, has been widely used in clinical application. With the rapid development of medical technology, people’s demand for information in medical field is more and more, but the traditional Nyquist sampling theorem has not been able to meet the demand, so it is urgently needed a new data acquisition and processing theory to deal with the pressure. In addition, as X-ray will cause harm to the human body, reducing the radiation dose has been the core of studies. Generally considering the two solutions:one is to reduce the transmission power of the X-ray source, which requires the improvement of detector’s performance, otherwise it would directly affect the image reconstruction quality; two is to reduce the sampling rate of projection data to shorten scan time. The second method is more convenient, namely to explore the image reconstruction method using sparse data.Therefore, this paper applies the theory of compressive sensing (CS) which emerged in recent years in order to reconstruct the medical image with high definition and complete details under the circumstance of sparse projection.This study is based on the iterative reconstruction algorithm on the previous research. It is aimed at solving the stripe artifact in simultaneous algebraic reconstruction technique (SART). The main research results are summarized as the following two aspects:1. An SART+TV algorithm based on compressive sensing. In this paper, on the basis of the theory of CS, this algorithm effectively removes the stripe artifact in SART in medical CT image reconstruction by using the total variation as constraint condition and the fast grid traversal method.2. An SART+K-SVD algorithm based on redundant dictionary learning. It uses prior images to train K-SVD redundant dictionary based on CS and then utilizes the dictionary to reconstruct the image after SART. The method can improve the quality of the reconstructed image obviously; furthermore prior training images can adopt a large number of CT images from the same section of the human body. The K-SVD dictionary can be preserved after once training, so that it can be used to reconstruct the CT image in the same section of the human body without training again. In this way the steps of medical CT image reconstruction are simplified.