Fig 1.
The moving blocker CBCT system.
The rotational axis of the gantry is z and the blocker is moving back-and-forth in z direction. The blocked region detects the scatter signal and the unblocked region detects the total signal, i.e. the sum of the primary and scatter signals.
Fig 2.
The process of the moving blocker CBCT reconstruction with scatter correction.
Fig 3.
The boundary effect due to the interpolation and extrapolation using limited sampling points in the single view SC (SVSC) method for the Catphan phantom.
Fig 4.
Blocker edge detection for a small object (left) and a large object (right).
The green arrows denote the position for reliable edge point detection in high-contrast regions. The edge is estimated as the line connecting two edge points. The red solid lines represent the true blocker edges and the red dashed lines are the estimated edges.
Fig 5.
Multi-view scatter estimation and adaptive threshold for blocker edge detection.
(a)-(f): the procedure to get the scatter profile (vertical axis) using the lowest signal in the signal profiles along the axial direction (horizontal axis) from multiple adjacent views (4 in this case); and (g) adaptive threshold (dashed red line) is derived from the scatter profile (red line) for more accurate and robust blocker edge detection, while the fixed threshold (green dashed line) fails to detect some edges in this case.
Fig 6.
The comparison of scatter signal estimated from the single-view scatter correction (SVSC) method (green line) and the multi-view scatter correction (MVSC) method (red line) for the Catphan phantom (3.2 mm pitch blocker).
The observed total raw signal is in black.
Fig 7.
Projection images of the Catphan phantom (the line integral of attenuation coefficients by taking logarithm of inversely normalized X-ray fluence): a) without blocker; b) with SVSC; and c) with MVSC. Each image was normalized to its maximum intensity value, respectively.
Fig 8.
Reconstructed images for CBCT of the Catphan phantom (close to the axial left boundary in Fig 7): (a, e) benchmark fan-beam MDCT; (b, f) without SC; (c, g) SVSC; and (d, h) MVSC. The red dashed line in (e) labels the transverse slice location. The difference between without SC and MVSC can be better discerned through the line profiles in Fig 9. (Display window [−900, 500] HU).
Fig 9.
Line profile through the middle line of the transverse slice in Fig 8.
The middle part is air and the CT number should be around -1000.
Fig 10.
The comparison of the scatter signal profiles estimated from SVSC (green line) and MVSC (red line) when the left marginal region is lack of scatter samples for the anthropomorphic pelvis phantom (9.6 mm pitch blocker).
The observed total signal is in black.
Fig 11.
Edge effect: Two transverse slices (in the top and middle row) and one coronal slice (in the bottom row) of the anthropomorphic pelvis phantom.
First column (a, e, i): MDCT; second column (b, f, j): CBCT without SC; third column (c, g, k): CBCT with SVSC; fourth column (d, f, l): CBCT with MVSC. The locations of the transverse slices are labeled in (i) (1 for the top row and 2 for the middle row). Regions of interest (ROIs) labeled in (e) were used for measuring the reconstruction accuracy quantitatively. (Display window [−600, 800] HU).
Table 1.
Comparison of the CT number (HU) of three ROIs of the Pelvis phantom in Fig 11.
Fig 12.
The comparison of scatter signal profiles estimated from SVSC (solid green lines) and MVSC (solid red lines) when the blocker edge detection error occurs for the anthropomorphic pelvis phantom (3.2 mm pitch blocker).
The red dashed line and green dashed line are the derived adaptive threshold and the fixed threshold (mean of the profile) for blocker detection, respectively. The observed total signal is in black.
Fig 13.
Robustness: Two transverse slices (in the top and middle row) and one coronal slice (in the bottom row) of the anthropomorphic pelvis phantom.
First column (a, d, g): MDCT; second column (b, e, h): CBCT with SVSC; third column (c, f, i): CBCT with MVSC of five adjacent views. The locations of the transverse slices are labeled in (g) (1 for the top row and 2 for the middle row). (Display widow [−600, 800] HU).
Fig 14.
Comparison between the proposed MVSC method and the clinical SC methods (U-SC: Uniform scatter correction; K-SC: Kernel-based scatter correction).
Fig 15.
Comparison of the horizontal line profile (red line in Fig 14) for Elekta with U-SC, Varian with K-SC and MVSC.
The MDCT profile serves as the ground truth.