Comparison of noise-magnitude and noise-texture across two generations of iterative reconstruction algorithms from three manufacturers

Diagn Interv Imaging. 2019 Jul-Aug;100(7-8):401-410. doi: 10.1016/j.diii.2019.04.006. Epub 2019 May 23.

Abstract

Purpose: To compare the noise-magnitude and noise-texture across two generations of iterative reconstruction (IR) algorithms proposed by three manufacturers according to the dose level.

Materials and methods: Five computed tomography (CT) systems equipped with two generations of IR algorithms (hybrid/statistical IR [H/SIR] or full/partial model-based IR [MBIR]) were compared. Acquisitions on Catphan 600 phantom were performed at 120kV and three dose levels (3-, 7- and 12-mGy). Raw data were reconstructed using standard "soft tissue" kernel for filtered back projection and one iterative level of two generations of IR algorithms. Contrast to-noise-ratio (CNR) was computed using three regions of interest: two of them placed in the low-density polyethylene (LDPE) and Teflon® inserts and another placed on the solid water. Noise power spectrum (NPS) was computed to assess the noise-magnitude (NPS peak) and noise-texture (NPS spatial frequency).

Results: CNR increased significantly in MBIR compared to H/SIR algorithms for General-Electric (GE) Healthcare (45%±12 [SD]) and Philips Healthcare systems (62%±11 [SD]) (P<0.001). Regarding Siemens Healthineers systems, CNR of MBIR was significantly lower than that of H/SIR (mean difference: -4%±5 [SD]) (P<0.001) for Teflon® insert but not for LDPE insert (mean difference: -4%±7 [SD]) (P=N.S.). NPS peaks were lower with MBIR than with H/SIR for GE Healthcare (-42%±8 [SD]) and Philips Healthcare (-75%±4 [SD]) systems, whereas it was greater with MBIR than with H/SIR for Siemens Healthineers (13%±11 [SD]) systems. NPS spatial frequencies were higher with MBIR than with H/SIR for Siemens (14%±10 [SD]) but lower for others (-17%±5 [SD] for GE Healthineers and -55%±3 [SD] for Philips Healthcare systems).

Conclusion: This study demonstrates that recent MBIR algorithms, by comparison with the preceding generation, differ according to the main manufacturers with respect to noise-magnitude and noise-texture.

Keywords: Image quality enhancement; Iterative reconstruction; Multidetector computed tomography; Noise power spectrum; Optimization.

Publication types

  • Comparative Study

MeSH terms

  • Algorithms*
  • Humans
  • Image Processing, Computer-Assisted*
  • Models, Statistical
  • Phantoms, Imaging
  • Signal-To-Noise Ratio*
  • Tomography, X-Ray Computed*