MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Virtual Acquisition: Efficient, Accurate and Low-Dose KV-Projection Based Positioning

H Yan1*, J Li1, Z Wang1, C Deng2, X Li2, C Ma3, J Li1, (1) Cyber Medical Research And Innovation Center, Our United Corporation,(2) Department Of Biomedical Engineering, Beijing Institute Of Technology,(3) Fox Chase Cancer Center, Philadelphia, PA,

Presentations

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: exists between DRR and its corresponding kV projection, introducing in-negligible discrepancy in 2D-3D/6D registration. This work is to fundamentally understand and solve the inconsistence problem, to enhance the kV-projection based positioning being with better accuracy, higher efficiency and lower-dose.

Methods: have developed a Virtual Acquisition (VA) algorithm. In each sub-cycle of VA, instead of physically moving/rotating the objects and re-acquiring kV projections, we reversely apply the registration results to the reference images and regenerate the DRRs. VA converges with consistent DRR and kV pair, and the result of all the sub-cycle are accumulated as the final registration output.

Results: becomes more significant when 1) the distance (R) from region-of-interest (ROI) to ISO is further, and 2) the setup position offset (r) is larger, especially with rotation (A). For a dot-structure at R=70 or 110mm, the inconsistence is calculated as 0.52 or 0.82mm when r=5mm. It increases to 1.04 or 1.63mm when r=10mm, and 1.15 or 1.92 mm with rotation (r=10mm, A=1deg). We conducted 2D-3D registration tests using an inhouse-designed box phantom, with 0-rotation and given offsets of 15mm. By applying VA, the mean/variance of the registration bias is decreased from {0.48, 0.33, 1.18}/{0.11, 0.06, 0.07} to {0.23, 0.18, 0.30}/{0.00, 0.00, 0.03} in mm, respectively. To compare VA based 2D-6D registration and CBCT based 3D-6D registration, we conducted tests on CIRS head phantom. The mean/variance of the differences regarding offsets and rotations are {0.30, 0.17, 0.36}/{0.01, 0.01, 0.01} in mm and {0.16, 0.23, 0.17}/{0.02, 0.06, 0.03} in degree. The extra computation time of VA is below 10 seconds.

Conclusion: Inconsistence between DRR and kV-projection could significantly affect the registration accuracy. VA algorithm can fundamentally solve the problem. With this algorithm, kV-projection based positioning truly becomes an efficient, accurate and low-dose image-guided patient setup approach.

Keywords

Image Guidance, DRRs, Image-guided Therapy

Taxonomy

IM/TH- Image Registration Techniques: 2D to 3D Registration

Contact Email