Impact of various iterative reconstruction algorithms on the noise power spectrum in low-dose CT imaging

The utilization of a low dose computed tomography (LDCT) protocol can pose potential drawbacks in image quality by increasing image noise. The purpose of this study was to assess the efficacy of iterative reconstruction (IR) for noise reduction by evaluating the noise power spectrum (NPS) reduction in LDCT CT brain imaging using a phantom study. This study involved scanning the uniform water section of the scanner phantom under different CT dose protocols. Reconstructions were performed using various slice thicknesses and vendor-specific IR algorithms. Three CT scanners of different vendors: Siemens, Philips, and GE were used. Two dose protocols for brain CT were used, the LDCT protocol with volume computed tomography dose index (CTDI vol ) of 35 mGy, and the standard protocol (CTDI vol of 65 mGy). The images were reconstructed using various IR algorithms such as Philips intelligence dose Level 4 (iDose 4 ) algorithm, Siemens sinogram affirmed iterative reconstruction (SAFIRE) Level 4 (S4) algorithm, and GE adaptive statistical iterative reconstruction (ASiR-V) algorithm and compared with filtered back projection (FBP). IndoQCT software was used to automatically calculate the NPS and noise level. The results show that the higher dose protocol and thicker slice acquisition reduce the noise level. Various IR algorithms reduced image noise and shifted the NPS curve down in comparison with filtered-back projection. The percentage of noise reduction in the LDCT protocol, compared to the reference image, was 12.82 %, 27.75 %, and 46.62 % for SAFIRE, iDose 4 , and ASiR-V, respectively, for 2 mm/2.5 mm slice thickness. For 5 mm slice thickness, the percentages of noise reduction were 10.95 %, 24.47 %, and 48.75 % for SAFIRE, iDose 4 , and ASiR-V, respectively. It can be concluded that IR algorithms proved to be effective in reducing noise in LDCT CT protocols, thereby enabling the generation of images with comparable quality to those obtained through standard dose (SDCT) protocols. • Vendor-specific iterative reconstruction (IR) algorithms effectively reduce noise and shift the noise power spectrum (NPS), but their performance varies by implementation. • IR techniques combined with thicker slice in low-dose CT offers image quality comparable to standard-dose CT. • Clinical application of IR must be carefully tailored, especially in high-resolution CT imaging procedures.