Detector Position and Gain Fluctuations Between Cone-Beam Computed Tomography (CBCT) Scanning
X Wang*, K Chao, S Wang, R Bassalow, J Chang, New York Weill Cornell Medical Ctr, New York, AASU-D-116-7 Sunday 2:05PM - 3:00PM Room: 116
To analyze the detector position and gain fluctuations between CBCT scanning modes using the scale invariant feature transform (SIFT) method.
A commercial on-boared imager was used to collect projection images of a humanoid torso phantom and a Catphan 600 Quality Assurance phantom for 360 degree gantry rotation. Two CBCT imaging protocols were investigated: (A) the 'full-fan' mode, i.e., the detector was fixed at the center position, and (B) the 'half-fan' mode with the detector shifted 148mm toward the right side of the phantom. For each gantry angle of the 'full-fan' collection, a corresponding projection image was interpolated for the 'half-fan' collection using images of two neighboring gantry angles. The SIFT method was used to detect the shape features in the overlapping region of the projection images for the same gantry angle from each collection. The results were two groups of matching feature coordinates, from which two dimensional (2D) rigid (translation and rotation) transform matrices were derived. The horizontal shift, vertical shifts and rotation angles between the 'full-fan' and 'half-fan' collections were calculated as a function of gantry angle. The gain difference between the two modes were also estimated using the SIFT matching points.
Using the 'full-fan' as the reference, the horizontal shifts, vertical shifts and rotation angles were 0.46±0.51 pixels, 0.29±0.34 pixels and 0.069±0.25 degrees for the torso phantom; and 0.25±3.38 pixels, 0.50±1.23 pixels 0.089±0.35 degrees for the CatPhan phantom. A systematic 3-4% gain fluctuation was also observed.
There are small but detectable detector position differences and systematic gain variations between CBCT scanning modes, which might cause artifacts for panoramic CBCT that stitches together the projection images of different views to expand the reconstruction volume. The SIFT method can also be used as a software stitching tool to correct these errors.
Funding Support, Disclosures, and Conflict of Interest: This work was partially supported by a DOD grant DOD W81XWH1010862