Skip to main content
. 2022 Jan 5;29(2):339–352. doi: 10.1007/s10140-021-02012-2

Table 2.

Comparison of quantitative image quality parameters for the noncontrast head CT series between ASiR-V and all three DLIR strength levels

ASiR-V DLIR-L DLIR-M DLIR-H P value
CT Attenuation (HU)
Thalamic GM 33.77 ± 2.44* 34.13 ± 2.18 34.05 ± 2.03 33.95 ± 2.02 0.001
PLIC WM 24.69 ± 2.30*ab 25.33 ± 2.03b 25.43 ± 1.99b 25.64 ± 1.99  < 0.001
M5 GM 36.99 ± 2.10ab 36.63 ± 2.04b 36.50 ± 2.00b 36.22 ± 1.97  < 0.001
CSO WM 26.86 ± 2.17b 27.04 ± 2.00 27.08 ± 1.88 27.17 ± 1.81 0.002
Image noise (HU)
Thalamic GM 5.51 ± 1.16ab 5.32 ± 1.11ab 4.39 ± 0.92b 3.46 ± 0.71  < 0.001
PLIC WM 5.74 ± 1.25ab 5.33 ± 1.02ab 4.31 ± 0.85b 3.23 ± 0.62  < 0.001
M5 GM 5.05 ± 1.04ab 4.87 ± 1.01ab 4.04 ± 0.83b 3.22 ± 0.72  < 0.001
CSO WM 5.46 ± 1.04*ab 5.10 ± 0.97ab 4.07 ± 0.78b 3.07 ± 0.58  < 0.001
PF (artifacts) 8.85 ± 1.39*ab 8.24 ± 1.02ab 7.12 ± 0.99b 5.94 ± 0.99  < 0.001
Air 6.83 ± 0.92*ab 4.75 ± 0.70ab 3.55 ± 0.60b 2.29 ± 0.50  < 0.001
ICH 6.01 ± 1.58ab 5.89 ± 1.62ab 5.00 ± 1.46b 4.23 ± 1.37  < 0.001
SNR
Thalamic GM 6.40 ± 1.37ab 6.69 ± 1.43ab 8.09 ± 1.68b 10.22 ± 2.11  < 0.001
PLIC WM 4.50 ± 1.05*ab 4.93 ± 1.02ab 6.13 ± 1.29b 8.23 ± 1.68  < 0.001
M5 GM 7.61 ± 1.51ab 7.81 ± 1.52ab 9.38 ± 1.79b 11.74 ± 2.39  < 0.001
CSO WM 5.11 ± 1.11*ab 5.50 ± 1.14ab 6.89 ± 1.36b 9.15 ± 1.76  < 0.001
ICH 10.44 ± 3.75ab 10.73 ± 3.73ab 12.68 ± 4.55b 15.18 ± 5.72  < 0.001
CNR
Thalamic GM-PLIC WM 1.65 ± 0.47ab 1.69 ± 0.44ab 2.03 ± 0.51b 2.53 ± 0.60  < 0.001
M5 GM-CSO WM 1.95 ± 0.50ab 1.95 ± 0.47ab 2.35 ± 0.54b 2.91 ± 0.65  < 0.001

HU Hounsfield units, GM gray matter, WM white matter, PLIC posterior limb of the internal capsule, M5 M5 cortex region (lateral MCA territory) according to Alberta Stroke Program Early CT Score — ASPECTS, CSO centrum semiovale, PF posterior fossa, ICH intracranial hemorrhage, SNR signal-to-noise ratio, CNR contrast-to-noise ratio, ASiR-V adaptive statistical iterative reconstruction-Veo, DLIR-L deep learning–based image reconstruction low strength level, DLIR-M deep learning–based image reconstruction medium strength level, DLIR-H deep learning–based image reconstruction high strength level

Post hoc pairwise multiple comparisons procedure with the Dunn-Bonferroni test showed a statistically significant (P < 0.05) difference between means when compared with DLIR-L (*), DLIR-M (a), and DLIR-H (b)