Research And Implementation Of Low Dose Algorithms Based On Dental CT

Computed Tomography, CT, was the emerging technique in the 1970s. Based on attenuation characteristic of X-ray and multiple angles of projections, CT can obtain gray cross-sectional images of scanned objects. With decades of developments, CT has experienced several generations. Due to its advantages, such as fast imaging, high resolution etc., it has become the important measurement tool in fields like medicine and industrial fields. Each CT breakthrough accompanies with the progress of industrial technology. In addition to the hardware development, reconstruction algorithm has been the research focus in CT field.Although computed tomography has become one of essential tools in doctors’daily diagnosis, CT dose is always a problem. There are many dose reduction methods, and promotion in the hardware is one of them. Besides, optimization of reconstruction algorithm is also a scheme to achieve the purpose of reducing dose, such as reducing the needed projection data. Therefore, how to implement the optimization algorithm to reduce dose becomes the research emphasis in this thesis.In theory study, this thesis conduct the following researches. (1) As the gold standard, FDK algorithm has been researched. Plus, based on FDK, we discuss the feasibility of half scan algorithm in practical CT. (2) In view of the actual situation, truncated data usually exist in commercial CT. Hence, this thesis will discuss how we can detect the truncation projections. (3) This thesis introduced a popular Backproject Filter algorithm (BPF) for region of interest (ROI) reconstruction. It can reduce the number of projections, and then achieve the goal of lower doses.In practice, this thesis expounds how to integrate the several kinds of low dose methods in actual CT system. The main part is how to accelerate BPF algorithm, including solutions under different number of GPU. The whole process of the reconstruction image, 512×512×512 size, based on single GPU is within 12 seconds after optimization; The method based on dual GPU only needs 4.118 seconds in reconstruction part, and the time of whole process is less than 9 seconds.A large number of experiments show that three algorithms are reliable and they are feasible in actual commercial production. Based on actual dental CT, the optimized algorithm has been applied. Both of reconstruction speed and result are satisfied the doctors’standard. Hence, the scheme in our thesis has a promising prospect.