Research On The Image Artifacts Correction And Key Problems In Clinical Use For Cone Beam Computed Tomography

In the past decades, there has been extensiveresearchon the X-ray dynamicimaging and three dimensional imaging using flat panel detector. The development ofdynamic flat-panel detector imaging technique enables the imaging mode of conebeam computed tomography (CBCT), which can perform high resolution threedimensional imaging of anatomic structure. Meanwhile, it can be used for themulti-view fluoroscopic imaging mode, which makes it possible for the tracking ofmoving target and for the dynamic imaging of rapid contrast agent intake and washout.These characteristics make X-ray dynamic imaging and three dimensional imagingtechniques widely used in medical imaging research and clinical diagnosis andtreatment.As for the flat-panel detector based CBCT imaging, the accurate geometry ofX-ray source, detector and axis of rotation must be known. The geometricmisalignments can cause significant impact on the quality of reconstructed sectionalimages. In this dissertation, an algorithm for the calibration and correction of CBCTgeometric parameters and misalignments was proposed based on the camera modelformulation in computer vision. In this algorithm, firstly, based on the camera modelformulation, the positions of3D markers in the calibration phantom and thecorresponding2D projections in the fluoroscopic images were used to reproduce theCBCT imaging geometry at each projection angle, from which the geometricparameters and misalignments can be derived. Then, based on the camera imagingmodel, the CBCT imaging formulation was modified to compensate the geometricmisalignments. At last, the FDK reconstruction algorithm was improved using themodified imaging formulation to reconstruct the sectional images from misalignedCBCT projections. In this algorithm, a general theory was proposed to describe thesystematic and complete calibration and correction for the CBCT geometricparameters and misalignments. The reconstructed image quality was improved usingthis method. This method has already been employed in our in-house dual-modalitysystem for small animal imaging.With the dynamic imaging capability of flat-panel detector, it became possiblefor the image guided pulmonary and abdominal tumor treatment in daily clinical use.However, in the guidance procedure using dynamic fluoroscopic images, the overlapping of the soft tissue causes the poor visibility of targets. Moreover, therespiration brings more difficulties for the localization of tumor position, whichmakes it harder for the accurate tracking of tumor motion. In this dissertation, thecapability volumetric and fluoroscopic imaging in CBCT system was used toco-register the high resolution3D volume and dynamic2D fluoroscopy based on thecamera imaging model. The techniques for tumor motion tracking and tumor visibilityenhancement were developed to provide augmented information for the image guidedtreatment.Using the X-ray dynamic imaging and CBCT imaging capabilities, we proposeda method to automatically track the tumor motion in image guided lung tumorradiotherapy. In this method, the respiration phase information was extracted from thefluoroscopic image at each projection angle and was used for the subsequentphase-resorting4D CBCT reconstruction. The4D CBCT volumes were integratedwith the3D CBCT volume to improve the image quality. After this, a2D-4Dmatching method was implemented to track the tumor motion in dynamicfluoroscopic images. To enhance the visibility of the tumor and surrounding tissues inimage guided hepatic needle targeting, we proposed a method to extract the interestedregions and targets from the3DCBCT volume and projected the segmented resultsonto the real-time fluoroscopic images. In this way, the visibility of the tumor andsurrounding tissues was enhanced to provide intuitive guidance for the hepatic needletargeting during free breathing.