基于散射核的锥束CT散射校正

发布时间：2018-10-19 18:58

【摘要】：锥束CT(CBCT)作为近年来发展的医疗及工业检测仪器,常用于图像引导治疗、上腹部检查、口腔3D重建、工业检测等方面。在CBCT成像系统中,CBCT重建的图像会受到X射线散射的严重破坏,大大降低图像的视觉效果。因此,许多研究学者开始把CBCT成像过程中的散射校正作为他们今后研究的重点问题,也成为目前的热点问题之一。而本文就该热点问题做了进一步的研究。首先,研究了CBCT因散射而形成的伪影及现存的CBCT常见重建方法。依据伪影产生的原因和特点,可以将CBCT图像中的伪影分为金属伪影、杯状伪影、散射伪影、运动伪影和硬化伪影等。在口腔和放射科中,金属伪影、散射伪影和杯状伪影是比较常见的三种伪影,这些伪影严重影响着CBCT的成像质量。针对CBCT散射校正后的投影数据,本文也详细介绍了CBCT重建方法中的一些常见方法,如解析法中的滤波反投影法和FDK算法。其次,本文就散射伪影问题提出了一种CBCT图像散射校正方法-基于正则化的盲去卷积。该方法大致可以分为三个步骤,第一:对CBCT的投影图像做预处理。对CBCT投影图像进行预处理,本文主要运用两种滤波器。这样做方便了散射核即卷积核的估计,在降低图像噪声的同时也能够突出图像的边缘。第二:估计出本方法所需的散射核。在本步骤中,首先利用第一步得到的预处理后的CBCT投影图像,再利用正则化方法估计出散射核。第三:对CBCT投影图像进行非盲去卷积。该步骤中主要是利用TV正则化方法和第二步求出的散射核,再进行非盲去卷积的计算。本方法最关键的一步就是如何求散射核。在上述步骤中,为了求解散射核的估计模型和图像的非盲去卷积模型,本文利用了两种算法,一种是FISTA,另外一种是IRLS。该方法主要是针对散射给所得图像造成的降质现象,图像一旦降质则会给医者诊断患者的某种疾病带来困扰,延误疾病治疗。为了得到降质前的图像我们就需要对散射投影图像进行校正,利用上述方法就可达到图像校正的效果,而实验结果也证明该方法的有效性,实验结果表明图像的伪影由9.435%降低到3.133%,而图像的对比度由108.34提高到138.87,提高了27.7%。最后,由于肿瘤体积估计精度对基于PET的放射治疗规划非常重要,因此本文也介绍了PET(Positron Emission Tomography)图像重建过程中运用盲去卷积的作用效果,实验结果表明,图像的对比度可以有从26%到46%的改善,因此,可以得出结论,重建的PET图像的盲去卷积可以用于提高肿瘤划分的精度,而且该方法容易实现。
[Abstract]:Cone-beam CT (CBCT), as a medical and industrial testing instrument developed in recent years, is often used in image guided therapy, epigastric examination, oral 3D reconstruction, industrial detection and so on. In CBCT imaging system, the image reconstructed by CBCT will be seriously damaged by X-ray scattering, which greatly reduces the visual effect of the image. Therefore, many researchers begin to make the scattering correction in the process of CBCT imaging a key issue in their future research, and become one of the hot issues at present. This paper makes further research on this hot issue. Firstly, the artifacts of CBCT due to scattering and the existing methods of CBCT reconstruction are studied. According to the causes and characteristics of artifacts, the artifacts in CBCT images can be divided into metal artifacts, goblet artifacts, scattering artifacts, moving artifacts and hardening artifacts. Metal artifacts scattering artifacts and goblet artifacts are three common artifacts in oral and radiology which seriously affect the imaging quality of CBCT. For the projection data corrected by CBCT scattering, this paper also introduces in detail some common methods in CBCT reconstruction, such as filtering back projection method and FDK algorithm in analytical method. Secondly, a CBCT image scattering correction method based on regularization is proposed in this paper. The method can be divided into three steps: first, the projection image of CBCT is preprocessed. In order to preprocess the CBCT projection image, two kinds of filters are used in this paper. In this way, it is convenient to estimate the scattering kernel, that is, convolutional kernel, which can reduce the noise of the image and highlight the edge of the image at the same time. Second, the scattering nuclei required by this method are estimated. In this step, the pre-processed CBCT projection images obtained from the first step are first used, and then the scattering nuclei are estimated by regularization method. Third, non-blind deconvolution of CBCT projection image. In this step, the method of TV regularization and the scattering kernel of the second step are used to calculate the non-blind deconvolution. The most important step of this method is how to find the scattering nucleus. In the above steps, in order to solve the estimation model of scattering kernel and the non-blind deconvolution model of image, two algorithms are used in this paper, one is FISTA, and the other is IRLS.. This method is mainly aimed at the degradation of the image caused by scattering. Once the image is degraded, it will bring trouble to the doctor to diagnose a disease and delay the treatment of the disease. In order to get the pre-degraded image, we need to correct the scattering projection image, and we can get the effect of image correction by using the above method. The experimental results also show that the method is effective. The experimental results show that the artifact of the image is reduced from 9.435% to 3.133%, while the contrast of the image is raised from 108.34 to 138.87, which increases 27.775%. Finally, because the accuracy of tumor volume estimation is very important for radiotherapy planning based on PET, this paper also introduces the effect of blind deconvolution in the process of PET (Positron Emission Tomography) image reconstruction. The experimental results show that, The contrast of images can be improved from 26% to 46%. Therefore, it can be concluded that the blind deconvolution of reconstructed PET images can be used to improve the accuracy of tumor classification, and this method is easy to implement.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R197.39;TP391.41

【相似文献】