MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel
  1. Home
  2. Meeting Program
  3. All Sessions
  4. All Imaging SNAP Oral
  5. Use of An Alternating Optimization Scheme with a Non-Linear Conjugate Gradient Method for Reconstructing High Quality CBCT and Motion Models From Free Breathing CBCT Projections

Use of An Alternating Optimization Scheme with a Non-Linear Conjugate Gradient Method for Reconstructing High Quality CBCT and Motion Models From Free Breathing CBCT Projections

M Guo1,2*, G Chee1, D O'Connell1 , J Fu1 , K Singhrao1 , D Ruan1 , P Lee1 , D Low1 , J Zhao2 , J Lewis1 , (1) UCLA School Of Medicine, Los Angeles, California, (2) Shanghai Jiao Tong University, Shanghai

Presentations

(Sunday, 7/29/2018) 5:05 PM - 6:00 PM

Room: Room 207

Purpose: We propose an alternating optimization scheme using a non-linear conjugate gradient method for reconstructing a high quality CBCT and corresponding motion model “of the day�. This method uses CBCT projections and external surrogate measurements acquired with the patient in treatment position, and avoids the need for deformable image registration or grouping projections into multiple bins.

Methods: The problem was formulated as an optimization problem where the fidelity term was constructed with both the reference image and the motion model as unknowns. For reference image reconstruction, a group of projections near peak-exhale were used. The motion model was estimated by minimizing the cost function while considering the reference image as constant. A non-linear conjugate gradient (CG) method was used to minimize the cost function to estimate the updated motion model. Motion model updating was done on a piece-wise basis to avoid introducing errors caused by first-order differentiation. By alternating between these two parts, a final reference image and motion model were obtained.

Results: We tested our proposed method with the XCAT phantom, with resolution 206×206, pixel size 2 x 2 mm, and 600 time points representing motion during 60 s. The average HU difference between our generated images and the ground truth XCAT images was 49.5±99.5 HU, with mean error of 3.5%. When reducing the projection number by 50%, the HU error increased to 59.1±201.4 HU.

Conclusion:
Conclusion: We developed a novel method for reconstruction of a high quality CBCT reference image and motion model of a patient in treatment position. Our results show that accurate motion models and reference images could be obtained for these initial phantom experiments. Future work will include extending the method to prospectively acquired patient images.

Funding Support, Disclosures, and Conflict of Interest: Partially funder by Varian master research agreement

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email

Additional Links