Evaluation of the Iterative Model Reconstruction (IMR) Algorithm for Four-Dimensional Computed Tomography (4DCT)
Abstract
Purpose
This work investigates use of the IMR algorithm in 4DCT for potential dose reduction in patients.
Methods
Lung motion and CATPHAN (with polystyrene sleeve) phantoms were scanned using 4DCT at exposures ranging from 200-1200 mAs. 4DCT images were reconstructed using filtered back projection (FBP), the hybrid iterative method (iDose4), and IMR using three tissue (soft-tissue, routine, sharp) and level (1, 2, 3) settings. The contrast-to-noise ratio (CNR) was determined for 1% low-contrast and acrylic targets in the CATPHAN, and for a spherical lung lesion in the motion phantom. The spatial frequency at 50% modulation transfer function (F50%), and the CT numbers of sensitometry targets were also determined. A composite quality factor Q (CNR * F50%) was calculated.
Results
CT numbers of sensitometry targets were consistent across all reconstruction types within 5 CT numbers. Lung lesion CNR was the highest for IMR level 3 with soft-tissue setting, which also showed the largest slope with respect to mAs. F50% of IMR reconstructions with soft-tissue setting were comparable to standard reconstructions (FBP and iDose4), while both the sharp and routine settings provided higher F50%. The factor Q was higher for all IMR settings compared to standard reconstructions at all mAs levels. Based on the CNR metric of the motion phantom and the quality factor Q of the CATPHAN, patient dose could potentially be reduced fourfold and threefold respectively compared to standard reconstructions. However, the ring artifacts at the lower mAs (≤400 mAs) in the standard reconstruction are not mitigated by the IMR reconstruction. Liver lesion conspicuity is improved by IMR (level 2, sharp) compared to iDose4 and FBP.
Conclusion
IMR has the potential to improve CNR without compromising spatial resolution while reducing dose in 4DCT. Aggressive dose reduction based solely on objective metrics should be carefully evaluated for artifacts in images.