Evaluation Of Ct Image Quality Of Abdominal Pancreas,Portal Vein,Kidney And Renal Vein By Deep Learning Image Reconstruction

Objective:To evaluate the value of deep learning image reconstruction(DLIR)algorithm compared with filtering back projection reconstruction(FBP)and adaptive statistical iterative reconstruction-veo(ASIR-V)algorithm in improving the CT image quality of pancreas,portal vein,kidney and renal vein in dynamic contrast-enhanced abdominal scan.Methods:Ninty-six patients who underwent conventional abdominal CT enhancement in Qingdao Municipal Hospital affiliated to Qingdao University were collected retrospectively.The obtained images were reconstructed by deep learning image reconstruction(DLIR),filtering back projection algorithm(FBP)and adaptive statistical iterative reconstruction-veo(ASIR-V).The deep learning reconstruction algorithm used low,medium and high levels for image reconstruction,and the iterative reconstruction algorithm used 30% ASIR-V and 70% ASIR-V for image reconstruction.The six reconstruction algorithm images obtained from each part were subjectively evaluated by two senior attending physicians in radiology department using double-blind method.The evaluation content includes the image noise,contrast and fine structure display.The effects of different reconstruction algorithms on image quality were observed by measuring CT value,standard deviation(SD),signal-to-noise ratio(SNR)and contrast to noise ratio(CNR)of different reconstruction algorithms in each study site,and comparing the six reconstruction algorithms in pairs.Results:There were statistically significant differences in the subjective scores of DLIR-L,DLIR-M,DLIR-H,FBP,30% ASIR-V and 70% ASIR-V of pancreas,portal vein,kidney and renal vein respectively(P < 0.001).In subjective scores,the two physicians had good consistency,Kappa value ≥0.69.DLIR-M image has the highest subjective score,followed by DLIR-H image.The SD value,SNR value and CNR value of the six reconstruction methods for each part of the study showed statistically significant differences(P < 0.001),among which the SD value of DLIR-H image was the lowest,and the SNR value and CNR value were the highest.In the reconstructed images in this study,the image evaluation of each target organ has the same change trend.Taking pancreas body as an example,with the increase of DLIR reconstruction grade,the SD value of images decreased 22.43%,37.86% and 54.46% compared with 30% ASIR-V.Compared with 30% ASIR-V image,SNR value and CNR value improved 28.62%,60.42%,117.67%,27.86%,59.20%,112.93%,respectively.Subjective score was also higher than30% ASIR-V reconstruction method.The image display of 70% weight ASIR-V reconstruction method is too smooth,and the subjective feeling is inferior to DLIR reconstruction image.Conclusion:Compared with the traditional FBP and current ASIR-V,DLIR can not only significantly reduce the noise and artifact of the image,improve the SNR and CNR,improve the display of structural details,improve the overall image quality and increase the diagnostic confidence of doctors.Therefore,combining subjective and objective evaluation of images,DLIR can be used to improve the quality of low-contrast abdominal CT images,which has a good clinical application prospect.