Encrypted login | home

Program Information

John's Equation Based Consistency Condition and Incomplete Projection Restoration Upon Circular Orbit CBCT

no image available
J Ma

J Ma1, H Qi1 , S Wu1 , H Yan2 , Y Xu1* , L Zhou1* , (1) Southern Medical University, Guangzhou, Guangdong,China (2) Cyber Medical Corporation, Xi'an,China

Presentations

WE-AB-207A-2 (Wednesday, August 3, 2016) 7:30 AM - 9:30 AM Room: 207A


Purpose:
In transmitted X-ray tomography imaging, projections are sometimes incomplete due to a variety of reasons, such as geometry inaccuracy, defective detector cells, etc. To address this issue, we have derived a direct consistency condition based on John's Equation, and proposed a method to effectively restore incomplete projections based on this consistency condition.

Methods:
Through parameter substitutions, we have derived a direct consistency condition equation from John’s equation, in which the left side is only projection derivative of view and the right side is projection derivative of other geometrical parameters. Based on this consistency condition, a projection restoration method is proposed, which includes five steps: 1) Forward projecting reconstructed image and using linear interpolation to estimate the incomplete projections as the initial result; 2) Performing Fourier transform on the projections; 3) Restoring the incomplete frequency data using the consistency condition equation; 4) Performing inverse Fourier transform; 5) Repeat step 2)~4) until our criteria is met to terminate the iteration.

Results:
A beam-blocking-based scatter correction case and a bad-pixel correction case were used to demonstrate the efficacy and robustness of our restoration method. The mean absolute error (MAE), signal noise ratio (SNR) and mean square error (MSE) were employed as our evaluation metrics of the reconstructed images. For the scatter correction case, the MAE is reduced from 63.3% to 71.7% with 4 iterations. Compared with the existing Patch’s method, the MAE of our method is further reduced by 8.72%. For the bad-pixel case, the SNR of the reconstructed image by our method is increased from 13.49% to 21.48%, with the MSE being decreased by 45.95%, compared with linear interpolation method.

Conclusion:Our studies have demonstrated that our restoration method based on the new consistency condition could effectively restore the incomplete projections, especially for their high frequency component.


Contact Email: