Session Item

Purpose of this work is to evaluate the benefit in the scope of radiation therapy planning of IMR, a model-based iterative reconstruction algorithm newly released by Philips for Big Bore CT version 4.8 with the promise of significative noise reduction. Image quality improvements were assessed for both standard and 4D reconstructions.

Spatial resolution, noise and low contrast detectability were measured at 120 KV using a Catphan 604 phantom for the three levels of noise reduction (1, 2, 3) and the two image definitions (Soft Tissue, Routine) offered by IMR (FOV 26cm, slice width 2mm, pitch 0.813). These metrics were compared to the current first generation iterative algorithm used clinically (iDose). Standard deviation a in 20mm diameter circular ROI was used as noise indicator while MTF at 50% and 10% were calculated for spatial resolution assessment. The largest low contrast rod (nominal 1.0%) was used for contrast calculation.  HU to relative electron density stability was evaluated using a CIRS model 062 phantom. For 4D acquisitions (helical acquisition at low pitch), the same Catphan phantom was moving using a dynamic platform during X-ray following a periodic cos4 pattern (20mm amplitude, period 4s) in the superior-inferior direction. 

IMR exhibits higher spatial resolution and better low contrast detectability than iDose for the same exposure. Spatial resolution is nearly independent of noise level reduction and exposure and is respectively 1.35 and 1.1-times higher than iDose (level 4, filter B) for image definition routine and soft tissue. IMR1 show similar noise level than iDose (level 4) while IMR2 and 3 provide respectively 33% and 59% noise reduction in average leading to 1.5 and 2.4-times higher contrast. The benefit of this contrast restitution relative to iDose increases while slice thickness decreases (1.36 at 3mm slice thickness and 1.63 at 1mm).A similar low contrast detectability and an improved spatial resolution is achievable with a dose reduction factor of 4 with IMR2 routine and 5 with IMR3 routine.Significative low contrast detectability improvements are observed for 4D reconstruction (phase binning, bin width 10%) with Soft IMR level 2 or 3 image definition soft tissue, with 1% contrast now visible above CTDI of 14mGy. HU are not influenced by IMR level or image definition type and HU to mass density relationship is nearly identical to iDose which facilitates the clinical adoption for dose calculation.

For a given dose level, the knowledge-based reconstruction algorithm IMR can improve both spatial resolution and low contrast detectability in comparison to iDose, improving overall image quality. IMR advantages relative to iDose increase when decreasing slice thickness and exposure. IMR 2 or 3 Tissue image definition significantly improve low contrast detectability for 4DCT acquisitions, parameter of interest in the abdomen area. HU numbers do not depend on image definition and noise reduction level and are similar to iDose.