Preliminary Application Of Artificial Intelligence-based Image Optimization In "Double-Low" Coronary CTA

Objective:To investigate the effects of Artificial Intelligence(AI)-based image optimization technique on image quality of "double low"(low tube voltage,and low contrast media volume)coronary CTA(CCTA).Material and Methods:Sixty patients,who were referred to perfonn CCTA,were prospectively collected between May and June 2018 in Peking Union Medical College Hospital and were randomly divided into two groups.Group A was scanned with a low tube voltage of 80 kVp and a reduced contrast media volume of lopamiro at 0.7 mL/kg and group B was scanned with a standard 120 kVp tube voltage and an injection of 70 mL lopamiro.According to the different reconstruction methods,group A was divided into two subgroups.The images of group Al were reconstructed with iterative reconstruction(IR).IR and further AI-based image optimization were used in group A2.Group B was also reconstructed by IR.To evaluate image quality objectively,the mean attenuation of contrast-enhancement values,background noise,signal-to-noise ratio(SNR)and contrast-to-noise ratio(CNR)were measured and calculated in the region of interests(ROIs)of the aortic root(Ao),left main coronary artery(LM),left anterior descending branch(LAD),left circumflex branch(LCX)and right coronary artery(RCA),respectively.In addition,the subjective evaluation was performed by two radiologists using Likert 4 scale(1 for excellent and 4 for poor)to evaluate the image quality of coronary artery branches and segments.The estimated radiation dose in terms of CT dose index(CTDIvol),dose length product(DLP)and effective dose(ED)was recorded and compared between group A and group B.Results:Comparison results of image noise,SNR and CNR among the three groups were as follows:at Ao,the image noise in three groups was significantly reduced in order of group Al,group B and group A2(all P<0.001)and SNR and CNR were significantly increased in order of group A1,group B and group A2(all P<0.001);at the coronary branchesof LM,LAD,LCX and RCA,the image noise in group A1 was significantly lower than that in group A2 and B(all P<0.001)and SNR and CNR were significantly higher than that in group A2 and B(all P<0.05).There was no significant difference in noise and SNR between group A2 and group B(all P>0.05),but CNR in group A2 was significantly higher than that in group B(allP≤ 0.001).The subjective evaluation results of coronary segments showed that group A2 and group B were significantly better than group A1(P= 0.002,0.038).There was no significant difference between group A2 andgroupB(P= 0.543).The radiation dose indexes of CTDIvol,DLP and EDin group A were significantly lower than those in group B(all P<0.001).The ED was decreased by 70.36%.Meanwhile,the volume of contrast media in group A was reduced by 37.11%than that in group B.Conclusion:Compared with conventional scanning,"double-low" CCTA images optimized by AI technology not only effectively reduced the radiation dose and contrast media volume,but also ensured subjective and objective image quality.