Program Information
Optimization-Based Reconstruction for Low-Contrast C-Arm CBCT Imaging
D Xia*, Z Zhang, B Chen, E Y Sidky, X Pan, The University of Chicago, Chicago, IL
Presentations
SU-H3-GePD-I-6 (Sunday, July 30, 2017) 4:00 PM - 4:30 PM Room: Imaging ePoster Lounge
Purpose: To investigate and develop 3D C-arm CBCT reconstruction for suppressing image artifacts caused by high-contrast objects involved in neurological surgery and treatment, thus obtaining improved images for visualizing intracranial structures.
Methods: The image to be reconstructed was formulated as a solution to a convex optimization program in which a data-derivative fidelity was introduced in the optimization program with an image-total-variation (image-TV) constraint for effectively suppressing image artifacts caused by high-contrast objects involved in the clinical procedures, and the Chambolle – Pock (CP) algorithm was tailored to reconstruct an image through solving the optimization program. In this work, a phantom study was conducted for image reconstruction from data collected by use of a clinical short-scan configuration with a physical head phantom. The reconstruction robustness was explored with respect to different data conditions with different high-contrast objects of practical relevance, including data acquired with a single high-contrast object at different positions and data acquired with multiple high-contrast objects. Quantitative metric, contrast-to-noise ratio (CNR), was used to evaluate the reconstruction quality of the physical head-phantom.
Results: Results show that the algorithm developed can significantly reduce streak artifacts observed in clinical FDK reconstructions and that low-contrast structures in the reconstruction can be more discernable than those in the clinical FDK reconstruction. This observation is corroborated by the quantitative CNR results.
Conclusion: The study demonstrates the effectiveness of the optimization-based reconstruction algorithm for artifact reduction as well as its potential to improve the visualization of low-contrast objects in CBCT imaging. Insights gained and algorithm developed in the study may be exploited for image reconstruction with reduced streak artifacts from the clinical patient data collected in interventional and surgical procedures.
Contact Email: