Click here to

Session: Cone-Beam CT [Return to Session]

Perturbation-Based Quality Surrogate for Scatter Estimation and Parameter Optimization in Anti-Scatter Grid (ASG)-CBCT

D Ruan1*, F Bayat2, C Altunbas2, (1) UCLA School of Medicine, Los Angeles, CA, (2) University of Colorado School of Medicine, Aurora, CO


SU-H300-IePD-F8-6 (Sunday, 7/10/2022) 3:00 PM - 3:30 PM [Eastern Time (GMT-4)]

Exhibit Hall | Forum 8

Purpose: The use of anti-scatter grid in cone-beam CT development promises improved scatter estimation and correction, but it is challenged by (1) the demand to reconstruct full spatial scatter distribution based on sparse samples, and (2) the lack of ground-truth to evaluate and optimize scatter estimation protocols. This addresses these unmet needs and applies the development to comparison of multiple scatter correction scheme.

Methods: At the center of our rationale is a novel perturbation-based scatter quality metric. Specially, without direct access to an absolute ground-truth scatter or the object in cone-beam HU, we design a surrogate metric to reflect physical feasibility: Both the scatter and object projection components are spatially smooth, with the former being homogeneous and the latter piecewise. We propose to use Tikhonov and total variation forms to reflect these properties, reflectively. To alleviate the impact of local geometry and sampling pattern from anti-scatter grid, a coarser-level perturbation is performed to make up for the myopic local gradient operations in classic Tikhonov and TV. Examination regions and perturbation windows are also designed to be conscious of such engineering designs.

Results: The proposed method has been implemented and applied to data acquired with a current ASG-CBCT system. It captures the behavior difference from various interpolation/extrapolation approaches in generating scatter field estimate from sampled measurements, agreeing with theoretical understanding. It further demonstrates that systematic interpolation and extrapolation combined with hyperparameter optimization based on this metric helps to reveal the object characteristics.

Conclusion: The proposed method provides a novel approach to define quality metric in the absence of ground-truth. It has been applied to competing method selection and hyperparameter tuning in the current context, and will be used further in image reconstruction tasks. The design idea is generalizable to a large range of developments when benchmark truth is not accessible.

Funding Support, Disclosures, and Conflict of Interest: This work is supported in part by NIH R01CA245270.


Not Applicable / None Entered.


IM- Cone Beam CT: Development (New Technology and Techniques)

Contact Email