Image quality assessment of pediatric chest and abdomen CT by deep learning reconstruction

Yoon, Haesung; Kim, Jisoo; Lim, Hyun Ji; Lee, Mi-Jung

doi:10.1186/s12880-021-00677-2

BMC Medical Imaging

Table 1 Quantitative image analysis of pediatric CT with different reconstruction techniques in comparison with 50% ASIR-V

From: Image quality assessment of pediatric chest and abdomen CT by deep learning reconstruction

Parameters		ASIR-V 50	ASIR-V 100		DLR-M		DLR-H
Chest CT without contrast enhancement (n = 16)
Attenuation (HU)	lung	− 795.9 ± 91.1	− 796.2 ± 94.2	0.872	− 797.6 ± 94.3	0.872	− 796.7 ± 93.9	0.872
Attenuation (HU)	paraspinal muscle	56.1 ± 11.1	54.6 ± 12.1	0.552	56.3 ± 10.7	0.552	54.9 ± 9.8	0.552
Noise		21.8 ± 3.7	15.6 ± 10.9	0.159	20.2 ± 3.8	0.455	14.6 ± 2.5	< 0.001
CNR	Lung	11.4 ± 3.7	22.9 ± 6.9	< 0.001	25.8 ± 13.3	< 0.001	28.4 ± 11.4	< 0.001
SNR	Lung	10.7 ± 3.5	21.4 ± 6.4	< 0.001	24.1 ± 12.5	< 0.001	26.6 ± 10.8	< 0.001
Chest CT with contrast enhancement (n = 12)
Attenuation (HU)	Lung	− 718.9 ± 139.4	− 718.7 ± 144.0	0.989	− 720.5 ± 145.4	0.989	− 719.1 ± 142.8	0.989
Attenuation (HU)	Paraspinal muscle	65.0 ± 8.5	64.6 ± 5.8	0.425	64.3 ± 5.2	0.425	62.6 ± 6.8	0.425
Noise		24.5 ± 6.1	12.6 ± 3.8	< 0.001	21.3 ± 4.9	0.172	14.8 ± 4.7	< 0.001
CNR	Lung	10.4 ± 4.0	17.8 ± 8.1	0.010	20.0 ± 6.9	0.001	21.4 ± 8.6	< 0.001
SNR	Lung	9.6 ± 3.7	16.4 ± 7.6	0.011	18.4 ± 6.6	0.001	19.8 ± 8.1	< 0.001
Abdomen CT with contrast enhancement (n = 23)
Attenuation (HU)	Liver	131.4 ± 28.4	131.7 ± 28.7	0.369	125.9 ± 39.1	0.369	132.6 ± 28.6	0.369
	Aorta	185.7 ± 45.2	184.9 ± 45.9	1.000	185.3 ± 45.1	0.227	187.6 ± 45.4	0.085
	Paraspinal muscle	71.2 ± 8.4	71.3 ± 6.6	1.000	72.7 ± 6.0	0.535	71.1 ± 6.2	1.000
Noise		19.9 ± 3.7	11.1 ± 3.6	< 0.001	16.3 ± 3.1	0.002	12.2 ± 2.4	< 0.001
CNR	Liver	3.2 ± 1.7	5.3 ± 3.0	< 0.001	3.2 ± 2.4	1.000	4.9 ± 2.5	< 0.001
CNR	Aorta	5.3 ± 2.2	9.5 ± 4.3	< 0.001	5.8 ± 1.9	0.012	8.0 ± 2.9	< 0.001
SNR	Liver	6.8 ± 2.1	11.5 ± 3.9	< 0.001	7.6 ± 2.7	0.583	10.7 ± 2.9	< 0.001
SNR	Aorta	8.5 ± 2.4	15.6 ± 5.1	< 0.001	9.7 ± 2.0	0.002	13.0 ± 3.2	< 0.001

The p-values < 0.05 were marked as bold
Values are presented as the mean ± standard deviation
ASIR-V 50, 50% adaptive statistical iterative reconstruction-V
ASIR-V 100 100% ASIR-V, DLR-M medium strength deep learning reconstruction, DLR-H, high strength DLR, CNR contrast to noise ratio, SNR signal to noise ratio. Image noise is based on standard deviation of paraspinal muscle attenuation

Back to article page

ISSN: 1471-2342

Contact us

General enquiries: journalsubmissions@springernature.com