Iterative Reconstruction Techniques for Radiation Therapy CT Simulations: A Phantom Study
H Li1*, D Lee1, D Low2, H Gay1, J Michalski1, S Mutic1, (1) Washington University School of Medicine, Saint Louis, MO, (2) UCLA, Los Angeles, CASU-E-J-167 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
To compare the Philips iDose⁴ iterative reconstruction techniques with filtered back-projection techniques (FBP) in Radiation Therapy and provide clinical practice insights.
An anthropomorphic pelvis phantom with added-bolus layers was used to mimic patients with 38-58 cm lateral diameters. For each phantom size, a set of CT scans were acquired on a Philips Brilliance 64-slice CT simulator with the mAs spanning the minimum to maximum dose. Images were reconstructed using FBP and iDose⁴ with noise reduction levels 1-6 for each scan, and evaluated on noise levels, contrast-to-noise ratios (CNR), CT number variations, and manual prostate contouring accuracy. Simulated prostate IMRT treatment plans based on these reconstructions were compared.
In general, greater-dose scans yielded greater CNR and lower noise, iDose⁴ reduced noise up to 66.1% and increased the CNR up to 53.2% compared to FBP. Only changing iDose⁴ noise reduction levels induced minor noise variations (0.21-4.83HU) and CT Number variations (<1 HU). Without changing reconstruction filters, the spatial resolution (represented by MTF), was similar on images reconstructed by FBP and iDose⁴. Very-low-dose scans yielded severe photon starvation artifacts, which decreased target visualization and could not be eliminated by iDose⁴, especially for the 58 cm phantom size. The gamma pass rates of the IMRT prostate treatment plans conducted using FBP or iDose⁴ based simulations were greater than 99.9% using 3%/3 mm criteria once they provide same segmentations as the ground truth.
Compared against FBP, iDose⁴ reduced noise, increased CNR and CT number consistency, produced pleasant images without losing structure details, but the improvements of target delineation confidence varied based on patient sizes and radiation output. Especially, very-low-dose scans (causing severe photon starvation artifacts) should be avoided when using iDose⁴ for obese patients. The iDose⁴ usage should be driven by target contour conspicuity and CT# accuracy instead of noise/CNR considerations.