深度学习图像重建在提高胸部CT图像质量的临床应用研究OACSTPCD

Objective To evaluate the effect of deep learning iterative reconstruction(DLIR)algorithm on the quality and noise of chest CT images.Methods The original data of 60 cases of chest CT examination in our hospital were retrospectively selected,and filtered back projection reconstruction(group A),adaptive statistical iterative reconstruction-veo 40%(group B),DLIR-medium(group C)and DLIR-high(group D).The CT values and SD values of lung tissue,aorta,muscle,thoracic vertebrae,and air were measured in four groups,and the signal to noise ratio(SNR)and contrast to noise ratio(CNR)of each tissue were calculated.Two physicians used a double-blind method to subjectively score image noise,artifacts,diagnostic confidence,and overall image quality on a 5-point scale.Results There was no significant difference in the CT values of lung tissue,aorta,muscle,thoracic spine and air between the four groups(P>0.05),but there was a statistically significant difference in noise values(P<0.05),and different reconstruction algorithms had a significant impact on the image noise.There were statistically significant differences in SNR and CNR among the four groups(P<0.05),among which the SNR and CNR of group D were the highest,and those of group A were the lowest.The subjective scores of the two physicians were in good agreement,and the Kappa value ranged from 0.781 to 0.884.There were significant differences in image noise,artifacts,diagnostic confidence,and overall image quality scores among the four groups(P<0.05),the subjective scores of groups C and D were higher than those of groups A and B,and there was no difference in subjective scores between groups C and D(P>0.05).Conclusion The DLIR algorithm can reduce the noise of chest CT images,provide higher image quality,enhance doctors'confidence in diagnosis,and has great potential to reduce radiation dose.