Encrypted login | home

Program Information

Optimization of Dual Energy Image Acquisition On the Gamma Knife Icon CBCT System for Improved Detection of Residual Gadolinium MRI Contrast Agent in Brain Tumors

no image available
V Grouza

V Grouza1*, SM Hashemi2, W Song1,2,3, A Sahgal2,3, Y Lee1,2,3, C Huynh1, H Nordstrom4, M Eriksson4, J Mainprize5, J Grafe1,2, M Ruschin1,2,3, (1) Department of Physics, Ryerson University, Toronto, ON, (2) Odette Cancer Centre, Sunnybrook Health Sciences Centre, Toronto, ON, Canada, (3) Department of Radiation Oncology, University of Toronto, Toronto, ON, Canada (4) Elekta AB, Stockholm, Sweden (5) Sunnybrook Research Institute, Toronto, ON, Canada


SU-E-201-7 (Sunday, July 30, 2017) 1:00 PM - 1:55 PM Room: 201

Purpose: To optimize dual energy (DE) image acquisition on the Gamma Knife Icon CBCT system for improved detection of residual gadolinium (Gd) MRI contrast agent in brain tumors.

Methods: DE-virtual monochromatic (VMC) images exhibit increased contrast of high atomic number (Z) elements. To evaluate the possibility of detecting trace amounts of Gd, we modeled realistic VMC image formation. Poly-energetic x-ray spectra for the Icon were simulated at 70 and 120kVp via a Monte Carlo platform provided by the vendor (based on Penelope). Energy-dependent noise characteristics incorporated Poisson statistics and were calibrated by comparing a virtual Catphan image with experimental measurements. Basis material (brain, bone) density maps of a head phantom with Gd-containing “tumors” (r=1cm, 100-1000ppm) were obtained by projection-domain DE-decomposition. VMC images were generated by reconstructing density maps at 51keV, the K-edge of Gd. Material separation was increased by hardening the 120kVp spectrum with filters of varied radiological thickness (0.1 – 2.0g·cm⁻²) and Z (40 – 60).

Results: Tumor contrast in simulated VMC images improved up to 150% following spectral filtration relative to unfiltered when compared with 120kVp scans at equal noise levels. Continuous improvement in contrast was observed for filters with Z up to ~50 and thickness <1.0 g·cm⁻². Increasing Z and thickness further yielded no additional improvement but rather increased image noise due to detrimental photon attenuation by the filter. Consequently, an example of an optimal filter includes tin (Z=50) of thickness 0.8g·cm⁻², which provides a ~140% contrast increase for a 10-fold decrease in photon fluence.

Conclusion: The low concentration of residual Gd following administration of MR contrast agent as well as high noise exhibited by CBCTs constitute a major challenge to the detection of brain tumors on the GK Icon. However, synthesis of VMC images from DE-CBCTs with additional filtration demonstrate promising results for improving tumor contrast.

Funding Support, Disclosures, and Conflict of Interest: This project was sponsored, in part, by Elekta AB, Stockholm, Sweden.

Contact Email: