Patient studies on shading correction for cone-beam computed tomography images
MetadataShow full item record
The work performed and presented in this thesis explores the efficacy of using planning Computed Tomography (CT) images as prior knowledge to improve quantitative cone-beam CT (CBCT) image quality in radiation therapy. CBCT is a significant component in the treatment planning process of image-guided radiation therapy (IGRT). Current CBCT images have various shading artifacts such as scatter, noise, and non-uniformity that create challenges in accurately identifying tissue abnormalities and reduce their usefulness for clinical applications. This thesis proposes a method to enhance the CBCT image quality when using commercial image correction methods (i.e. images corrected by a Varian algorithm). The results show that scattering and image non-uniformity are greatly reduced by the proposed method. Therefore, the proposed method achieves better image correction results than does the Varian correction algorithm. Since all patients who receive radiation treatment routinely undergo a multiple detector array CT (MDCT) scan as part of the diagnostic procedure, the high quality MDCT serves as the “free” planning CT (pCT). To improve the CBCT images that are taken during radiation treatment, the CBCT is first spatially registered with the pCT via rigid and deformable registration using Velocity software. Primary projections in the CBCT scan are estimated via forward projections of the registered MDCT image. The low frequency errors in the projections, which are a major cause of shading artifacts in CBCT images after reconstruction, are estimated by filtering the difference between the original line integral and the estimated scatter projections. The corrected CBCT image is then reconstructed from the projections using the Feldkamp, Davis, and Kress (FDK) algorithm. With the planning MDCT treated as ground truth, the CBCT image corrected by the proposed method is compared against the corrected image using the Varian Medical Systems (VMS) algorithm, a commonly used commercial shading correction method. The results are presented in the axial, coronal, and sagittal views, and are evaluated by comparing the mean number of error for three image quality evaluation factors - CT number, spatial non-uniformity (SNU) value, and image contrast value. A paired t-test is performed on the results to prove the consistency and reliability of the proposed method of shading correction. The proposed method is evaluated on 20 sets of thorax and pelvis cancer patient data from Cancer Treatment Centers of America (CTCA). CT numbers, measured in Hounsfield units (HU), for four uniformly selected regions of interests (ROIs) are found in each set of images. The mean errors in CT number, SNU, and contrast value for the Varian corrected image and the image corrected by the proposed method are 53 HU and 41 HU, 7.3% and 3.0%, and 37 HU and 34 HU respectively. The results show that as compared to the Varian correction algorithm the proposed method delivers a CBCT image quality with better spatial uniformity and fewer CT number errors at the 95% confidence level, and with statistically insignificant but comparable change in image contrast. Promising results have previously been obtained using similar methods on CBCT tabletop phantoms and a limited amount of prostate patient data sets. As an extension of previous work performed by Niu et al, the proposed method is modified and evaluated using 20 sets of thorax and pelvis cancer patient data, which includes both male and female patients. Statistical analysis is performed and confirms that the proposed method can be employed within the current CBCT shading correction algorithm used by VMS to greatly enhance CBCT image quality.