Deep Learning Reconstruction for Contrast Agent of Lower Concentration for Abdominal CT Between Filters Back Projection and Iterative Reconstruction on Improving Image Quality

The number of CT examination is increasing, and the exposure dose and side effects of contrast agents are also increasing. Recently, Many efforts have been made to solve the problems mentioned by the technological advances of artificial intelligence and deep learning. In this study, Deep learning reconstruction(DLR) AiCE quantitatively compared the imaging characteristics of a conventional filter back projection(FBP) and Hybrid iterative - reconstruction(H-IR) AIDR by concentration difference. The FBP, AIDR, and AiCE were applied to 3 phases(by concentration) in 10 of the 46 patients who underwent liver contrast examination in February 2023. The image quality was evaluated by concentration using Noise, SNR, and CNR. The average rate of change in noise was effective decrease of 67.45 %, 73.98 %, and 20.05 % in the order of AIDR compared to FBP, AiCE compared to FBP, and AiCE compared to AIDR. SNR increased by 144.84 %, 186.50 % and 17.01 %. CNR increased by 142.14 %, 182.38 %, and 16.53 %. Delay(128HU)-AiCE compared to Artery(337HU)- FBP in SNR, Portal(172HU)-AiCE compared to Artery(337HU)-FBP in CNR showed no statistically significant difference. When DLR is applied even at low HU values, improved image results are obtained, so it will be helpful for the use of low concentrations and small amounts of contrast agents.