Fig. 4
From: Deep learning-based metal artifact reduction in CT for total knee arthroplasty
Results of quantitative analysis from the clinical test dataset. (a) The area of dark streak artifacts was the smallest in the KMAR-Net, reducing most of the streak artifacts. (b) The mean HU within the dark streak artifacts was higher in KMAR-Net than in O-MAR. (c) The standard deviation (SD) within the artifacts was the lowest in KMAR-Net.
