Study On Iterative Reconstruction Algorithm For Region Of Interest Based On Two CT Scans

X-ray computed tomography(CT) is a technique to noninvasively acquire the inner structure information utilizing the attenuation of X-ray passing through objects. With the widely usage in medical diagnosis, the radiation dose problem of CT draws more attention. As the CT devices are all in full-field scan, there is an obvious problem of over radiation when the patient only needs to be scanned for certain tissue or organ. The method of combining low-dose coarse scan and local accurate scan to reconstruct the image is the efficient way to cut down the dose received by patients.To achieve the CT local iterative reconstruction combining low-dose scan and accurate scan, this thesis studies the following four works:1、Automatic Heart Positioning Method in CT Scout Images. It canobtain a three dimensional positioning of the heart region based on the anteroposterior scout image and lateral scout image. This is an essential step to perform the ROI scan is to automatically positioning the heart.2、Study the algebraic reconstruction technique(ART). It is based on solution of algebraic linear equations algorithm and the formula is derived from the least-squares method. Due to the property of anti-noise and strong robustness, the iterative reconstruction algorithm is more superior to the analytic algorithm in low-dose scan and can still reconstruct high-quality image.3、Generate the raw data of any low-dose scan from the raw data of accurate scan. Based on X-ray photons Poisson statistical distribution we form a statistic model and by changing the value of tube current it can generate the raw data of corresponding low dose scan.4、Use the raw data which combines the low-dose scan and accurate scan to realize local iterative reconstruction. By weighted combining the low-dose scan’s and accurate scan’s raw data, we can reconstruct high-quality image of region of interest through ART.5、Study the reconstruction algorithm which imports the smoothing process and is droved by pixels. Since the reconstruction image droved by rays is coarse and noisy, to smooth the image, we add the regularization term of considering neighboring pixels in order to improve the smoothness of image. Through the analysis and comparison of reconstruction image from the stimulated data, this method can smooth the image in a way.6、Registration of CT coarse scanning data and fine scanning databased on rigid transformation.Reconstruction will be not correct based on the two data sets since position of the patient and organs probably change in two scannings.Through the experiment, the reconstructionforregion of Interest based on two CT scans is gotten, and the reconstruction algorithm imports the smoothing process and is droved by pixels. In addition, registration of CT coarse scanning data and fine scanning data improves the reconstruction image to some extent.