Scatter correction for cone-beam computed tomography using moving blocker strips

One well-recognized challenge of cone-beam computed tomography (CBCT) is scatter contamination within the projection images. Scatter degrades the image quality by decreasing the contrast, introducing shading artifacts, and leading to inaccuracies in the reconstructed CT number (voxel density). Various strategies have been investigated to manage the scatter signal in CBCT projection data. Among them, blocker-based techniques have shown excellent results in suppressing scatter-related artifacts. However, current blocker-based methods may require an additional measurement for each projection view¹ or may reduce the imaging volume^2,3 corresponding to the blocked region in projection data.

We propose a moving blocker-based approach to simultaneously estimate scatter signal and reconstruct the complete volume within the field of view (FOV) from a single CBCT scan.⁴ The measured signal in the blocked region is from the scatter. In the unblocked region, the measured signal is the sum of the primary and scatter signals. The primary signal in the unblocked region is obtained by subtracting the estimated scatter signal from the measured one. Figure 1 illustrates the geometric setup of the CBCT imaging blocker, which consists of lead strips aligned along the detector's v-direction. The blocker is inserted between the x-ray source and the patient. It moves back-and-forth along the detector's u-direction as the gantry rotates around the z-axis. With this design, the data required for ramp-filtering in the Feldkamp-Davis-Kress (FDK) algorithm is complete in the unblocked region. The area within the FOV has measurements at different views. In the FDK's back-projection step, every voxel has contributions from the unblocked projection at different views. Therefore, our proposed strategy makes it possible to reconstruct the entire volume using the analytical FDK algorithm from a single CBCT scan. We further developed an iterative image construction algorithm based on constraint optimization to rebuild the CBCT images from the measured signal in the unblocked region following scatter correction.

Figure 1.The blocker is inserted between the x-ray source and the patient. It moves back and forth along the u-direction as the gantry rotates around the z-axis.

We performed simulation and experimental phantom studies to evaluate the proposed scatter-correction strategy. In the simulation study, the mean relative error was reduced from 25% to 3% and 2% in the images reconstructed by the modified FDK and constraint optimization, respectively. In the experimental study using a CatPhan 600 phantom performance measurement instrument, CT number errors in the selected regions of interest were reduced from 256 to fewer than 20. Figures 2 and 3 show CatPhan 600 images obtained by the different methods. Figure 2 shows axial images of the sensitometry module. Figure 3 shows the phantom's coronal view, demonstrating that the proposed strategy can reconstruct the entire FOV volume. We observed shading artifacts and non-uniformity in images without scatter correction—Figures 2(a) and 3(a)—whereas images were uniform and shading artifacts greatly suppressed in images using our proposed scatter-correction strategy. We used the modified FDK algorithm to construct images from partially-blocked projection data after the estimated scatter signal was subtracted from the measured signal: see Figures 2(b) and 3(b). While scatter-related artifacts were substantially suppressed, the noise level was much higher than that of images without scatter correction. Noise was greatly suppressed in the images after we processed projections by the penalized weighted least-squares algorithm⁵: see Figures 2(c) and 3(c). Figures 2(d) and 3(d) show images rebuilt using constraint optimization from projections after scatter correction. The image quality obtained by the iterative constraint optimization algorithm was improved substantially compared with the FDK analytical algorithm.

Figure 2. Axial slices of the CatPhan 600 sensitometry module with (a) FDK from scatter-contaminated, full-projection data, (b) modified FDK from partially blocked projection data where estimated scatter signal is subtracted, (c) modified FDK from scatter-corrected projection processed by the penalized weighted-least-squares (PWLS) algorithm, and (d) constraint optimization from partially blocked projection data where estimated scatter signal is subtracted. Display window [-600 400] Hounsfield units (HU).

Figure 3. Coronal views of the CatPhan 600 with (a) FDK from scatter-contaminated, full-projection data, (b) modified FDK from partially blocked projection data where estimated scatter signal is subtracted, (c) modified FDK from scatter-corrected projection processed by the PWLS algorithm; and (d) constraint optimization from partially blocked projection data where the estimated scatter signal is subtracted. Display window [-600 400] HU.

In summary, we have proposed an effective scatter correction scheme for CBCT. We inserted a moving blocker consisting of lead strips between the x-ray source and the patient during CBCT acquisition. The proposed method allows us to simultaneously estimate the projection data-scatter signal, reduce the imaging dose, and obtain complete FOV volumetric information.

We are evaluating this new scatter-correction scheme on more realistic anthropomorphic phantoms and patients. The technique can be used to improve target accuracy in image-guided radiation therapy as well as to enhance the diagnostic accuracy of flat-panel, detector-based imaging modalities such as breast CBCT and tomosynthesis.